University of Southern California Dissertations and Theses

Structural Factors Affecting Fertility In Large United States Cities

(USC Thesis Other)

Structural Factors Affecting Fertility In Large United States Cities

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content This dissertation has been
microiilmed exactly as received 70-8533
MARSHALL, Jr., Harvey Huston, 1939-
STRUCTURAL FACTORS AFFECTING
FERTILITY IN LARGE UNITED STATES CITIES.
University of Southern California, Ph.D., 1969
Sociology, general
University Microfilms, Inc., Ann Arbor, Michigan
STRUCTURAL FA CTORS AFFECTING FERTILITY IN
LA RG E UNITED STATES CITIES
by
Harvey Huston Marshall, J r .
A d isse rta tio n Presented to the
FA CU LTY OF THE G R A D U A TE SC H O O L
UNIVERSITY OF SO U TH ERN CALIFORNIA
In P artial Fulfillm ent of the
Requirements fo r the Degree
D O C TO R OF PHILOSOPHY
(Sociology)
August 1969
UNIVERSITY O F SO U T H E R N C A LIFO R N IA
THE GRADUATE SCHOOL
UNIVERSITY PARK
LOS ANGELES, CALIFORNIA 9 0 0 0 7
This dissertation, written by
m r v e y o t s t ^ ...........................
under the direction of h.is.... Dissertation Com
mittee, and approved by all its members, has
been presented to and accepted by The Gradu
ate School, in partial fulfillment of require
ments of the degree of
D O C T O R OF P H I L O S O P H Y
Dean
Date A u S U 3 t 1 9 6 9
DISSERTATION COMMITTEE
Chair pain
Llv.
A C K N O W LED G M EN TS
I would lik e to acknowledge the g re a t assistan ce given by
Drs. Sanford Labovitz and Jon M iller in the preparation of th is
d is s e rta tio n . Their s k ille d c ritic ism and suggestions were
invaluable.
ii
TABLE OF CONTENTS
Page
A C K N O W LED G M EN TS ......................................................................................................... i1
LIST OF TABLES......................................................................... vii
Chapter
I. STATEM ENT OF THE PROBLEM........................................................................ 1
Rural-Urban Differences
Size of Place
Human Ecology and F e r tility
General Perspectives
Summary and Objectives
I I . ECO N O M IC A N D SOCIAL DETERM INANTS OF FERTILITY ............................. 8
Economic Determinants
Business cycles and f e r t i l i t y
Income and f e r t i l i t y
Summary
Social Determinants
Labor force p artic ip a tio n of women
Education
Occupation and f e r t i l i t y
Social m obility and f e r t i l i t y
Horizontal m obility
Black f e r t i l i t y
I I I . TH EO RY A N D HYPOTHESES............................................................................ 35
S tructural Variables and Hypotheses
Income and f e r t i l i t y
Cost and f e r t i l i t y
Integration of women and f e r t i l i t y
Education and f e r t i l i t y
Bureaucratization
111
Chapter
Migration
Social m obility
Family stru ctu re
IV. SOURCES OF D A TA A N D M ET H O D S OF ANALYSIS
Sources of Data
Temporal sequence
Ecological correlations
Operational Measures of Variables
Income
Cost
Integration of women
Education
Migration
Social mobility
Bureaucratization
Occupational d iffe re n tia tio n
Segregation
Family disorganization
F e r tility
Methods of Analysis
Pearsonian product moment correlation
P artial correlation
M ultiple correlation and regression
V. ANALYSIS A N D DISCUSSION ..................................
Income and F e r tility
Total population
Nonwnite population
Cost and F e r tility
Total population
Nonwhite population
Integration of W om en and F e r tility
Total population
Nonwhite population
iv
Chapter
VI.
Education and F e rtility
Total population
Bureaucratization and F e r tility
Total population
Nonwhite population
Migration and F e rtility
Total population
Nonwhite population
Social Mobility and F e r tility
Total population
Nonwhite population
Family Disorganization and F e r tility
Total population
Nonwhite population
Segregation and F e r tility
Multiple Regression and Correlation Analysis
Total population
Nonwhite population
Discussion
S U M M A R Y A N D IMPLICATIONS ...............................................
Summary
Implications
Ecological prespectives
Rural-urban differences
F e rtility control
Limitations and Suggestions for Future Research
Causal order
Other variables
Page
, 116
v
Chapter Page
A dditivity assumption
Lack of rural data
Unit of study
United States settin g
REFERENCES.......................................................................................................... 129
vi
LIST OF TABLES
Table Page
1. Indexes and S o u rc e s ............................................................................... 56
2. Examples of Mobility contexts ......................................................... 64
3. Zero and P artial Correlation C oefficients Between
Measures of Income and F e r tility ........................................... 78
4. Zero and P artial C orrelation C oefficients Between
Measures of Education and F e r tility ....................................... 85
5. Zero-order and P artial Correlation C oefficients
Between Measures of Female Education and
Ferti 1 i t y ............................................................................................... 87
6. Zero-order and P artial C orrelation C oefficients
Between Bureaucratization and Measures of F e r tility . . 90
7. Zero-order and P artial Correlations Between
Measures of Migration and F e r tility ....................................... 95
8. Zero-order and P artial Correlations Between
Social Mobility and Measures of F e r tility .......................... 99
9. Zero-order and P artial Correlations Between
Family Disorganization and Measures of F e r tility . . . 101
10. Zero-order and P artial Correlations Between
Segregation and Measures of F e r tility ................................... 105
11. Standardized Regression C oefficients in Regression
Equations with the F e r tility Ratio as the
Dependent V ariable--Total Population ................................... 109
12. Standardized Regression C oefficients in Regression
Equations with the F e r tility Ratio as the
Dependent Variable—Nonwhite P o p u la tio n ................................... 112
v ii
CHAPTER I
STATEM ENT OF THE PR O B LEM
Sociologists usually approach the analysis of f e r t i l i t y from
two major perspectives—social psychological and s tru c tu ra l. In a
social psychological perspective emphasis is upon individual
c h a ra c te ristic s which a ffe c t f e r t i l i t y . In a stru ctu ra l perspective
analysis links changes in the properties of e n tire systems to
changes in f e r t i l i t y . In sociology a major emphasis of th is approach
has been the analysis of patterns of rural-urban f e r t i l i t y .
However, although emphasized, the possible re la tio n s between
urbanism and f e r t i l i t y have not been organized into the rigorously
defined se t of in te rre la te d hypotheses usually reserved fo r the term
"theory." In p a rtic u la r, the link between urbanism and f e r t i l i t y is
never c learly specified. That i s , what is i t about the c ity which
leads to changes in f e r t i l i t y ?
The principal goal of th is d isse rta tio n is explication of the
nature of the link between urbanism and f e r t i l i t y . The Importance of
such research is increasingly recognized as socio lo g ists become more
aware of the presently ambiguous and occasionally inconsistent
in terp reta tio n s of th is rela tio n sh ip . Some d irectio n is offered by
Stycos (1968), who has proposed th a t sociologists ask how urbanism is
rela ted to f e r t i l i t y . He notes th a t Latin America c i tie s d iffe r from
1
2
rural areas in income, labor force p articip a tio n of women, church
attendance, and many other variables. He asks how much variance
remains to be explained by something called "urbanism" i f these
variables are held constant.
Freedman (1961-1962) has made e sse n tia lly the same point 1n
his injunction th a t socio lo g ists explain what i t is about urban
environments th a t a ffe c t f e r t i l i t y . He says:
The question is whether community size and farm-urban
d iffe re n tia ls r e f le c t d iffe re n tia ls in other variables which
are changing ( e .g ., in s titu tio n a l sp e c ia liz a tio n , or the
extent to which economic or other a c tiv ity is based on a
fam ilial u n it or education le v e ls). This involves explain
ing why c ity size or rural-urban differences a ffe c t
f e r t i l i t y in the f i r s t place. (1961-1962:67)
Another important goal is comparison of the determinants of
black and to ta l f e r t i l i t y le v els. This phase of analysis is
prim arily inductive and comparative. Concern is with establishing
what d ifferen ces, i f any, e x ist between the two groups.
Rural-Urban Differences
That there is something about urban environments th a t leads to
low f e r t i l i t y is most clearly indicated by the existence of lower
f e r t i l i t y in these areas as compared with ru ra l. In f a c t, the
ubiquitous, nature of th is d iffe re n tia l has sometimes been taken as
conclusive evidence th a t c i tie s create low f e r t i l i t y . An early paper
by Jaffe (1942) is a useful point of departure. Jaffe examined then
current lite r a tu r e to ascertain the extent and importance of th is
d iffe re n tia l in a variety of geographical and cultural areas. He
concluded th a t rural urban d iffe re n tia ls were both widespread and of
3
long standing, and remarked th a t they were " . . . neither increasing
nor decreasing with the passage of time" (1942:57-58).
Regarding the current s itu a tio n , Stycos (1968) concluded th a t
there can be l i t t l e question about.the existence of rural-urban
differences in Latin America. The Population Branch of the United
Nations (1961) reached the same conclusions about th is area. A large
number of other scholars have documented th is pattern in many other
areas. (See, for example, Davis, 1951; Taueber, 1958.)
However, despite the general tendency fo r urban environments
to have lower f e r t i l i t y than ru ra l, a number of anomalies suggest the
need for fu rth er analysis. Robinson (1963) has suggested th a t .When
ever conclusions about d iffe re n tia ls are based upon child-woman ra tio s
(which they usually are) there is serious question as to th e ir
accuracy. Using a correlation analysis he showed th a t the association
between per cent urban and f e r t i l i t y is actually an in d ire c t measure
of per cent married and in fan t m ortality. This suggests th a t an
apparent rural-urban d iffe re n tia l is often only a measure of these
variables.
Furthermore, the Robinsons (1950) and Zarate (1967) have
raised the p o s sib ility th a t a recent decline in the magnitude of the
d iffe re n tia l in Mexico may be due to a ris e in urban f e r t i l i t y .
Collver (1965) has shown th a t Argentina, Chile, and Venezuela are
experiencing a rapid r is e in f e r t i l i t y and the percentage of persons
living in c itie s . Sim ilarly, i t has been carefu lly documented th at
rural f e r t i l i t y has fa lle n much more rapidly since the twenties than
4
has urban f e r t i l i t y in the United S tates. (See, fo r example,
E asterlin , 1962; and Okun, 1958.) F inally, not only did to ta l f e r t i l
ity ris e rapidly during the la te fo rtie s and through the la te f i f t i e s ,
urban f e r t i l i t y rose more rapidly than rural f e r t i l i t y .
Size of Place
In addition to the rural-urban d if f e r e n tia ls , the inverse
association between size of place and f e r t i l i t y is often cited as
evidence th a t urban environments produce low f e r t i l i t y . The im plicit
hypothesis is th a t larg er places exhibit more d is tin c tly urban
c h a ra c te ristic s. W hen th is relatio n sh ip is examined fo r aggregated
data i t appears valid . For example, Duncan and Reiss (1956) showed
th a t there is a consistent tendency for the f e r t i l i t y of populations
residing in larg er urbanized areas to be lower than th a t in the
smaller areas.
However, when th is relatio n sh ip is examined for central c itie s
i t tends to disappear. As early as 1931, Thompson (1931) observed
th a t size of central c ity seemed to have no consistent re la tio n with
f e r t i l i t y . Grabill and his associates (1958) came to esse n tia lly the
same conclusion in th e ir analysis of data for c itie s of 250,000 or
more in 1940. And f in a lly , fo r 1960, Hadden and Borgatta (1965) found
th a t the correlation between size of place and f e r t i l i t y fo r all
c i tie s in the United States to be -.01.
From th is review, i t is apparent th a t the data fo r both
rural-urban d iffe re n tia ls and size of place are often inconsistent.
While c itie s tend to have lower f e r t i l i t y than rural areas, th is is
5
not inevitably the case. Moreover, the f e r t i l i t y of the urban sector
has often risen while th a t of the rural sector f e l l . Finally, size
of place tends not to be related to f e r t i l i t y for central c itie s in
the United S tates.
Even i f the data were co n sisten t, demographers would s t i l l be
fa r from an explanation. Specification of some model linking the two
is e sse n tia l. However, given these inconsistencies demographers are
unable e ith e r to p red ict or explain f e r t i l i t y .
Human Ecology and F e r tility
A p art of the fa ilu re to explain the relationship between
urbanism and f e r t i l i t y is a ttrib u ta b le to inadequate conceptualization
of the problem. Sociologists have not placed th e ir studies of the
correlation between urbanism and f e r t i l i t y into a more general
theoretical perspective, and the definitions of these concepts are
rarely derived from any broader framework. A framework which will be
applied here to both of these problems is human ecology.
As explicated by Duncan (1959), human ecology is concerned
with the co llectiv e adaptation of a population to i t s environment.
This adaptation can be described in terms of four interdependent sets
of variables: population, organization, environment, and technology.
Because of th e ir interdependence changes in any one s e t has casual
im plications for changes in each of the others. This formulation is
generally consistent with other statements about the nature of human
ecology (cf. Gibbs and Martin, 1959).
Of p a rtic u la r importance in the present context are the
6
population and organization variables. In th is approach, f e r t i l i t y
is conceptualized as a dependent variable in the ecosystem complex.
Indeed, the argument im p lic it in most studies of urbanism and f e r t i l i t y
is th a t as the organization of a society changes from rural to urban,
various forces are s e t in motion which ultim ately re s u lt in lowered
f e r t i l i t y . Such an approach is inherently ecological. What remains
is explication of the nature of the organizational facto rs and th e ir
link to f e r t i l i t y .
General Perspectives
Human ecology provides a general framework, but i t does not
provide specific hypotheses about the dimensions of urban stru ctu re
and th e ir re la tio n to f e r t i l i t y . These w ill be derived from two
general perspectives. The f i r s t , id e n tifie d prim arily with economic
theory, suggests th a t f e r t i l i t y is prim arily the product of "cost"
factors in urban environments. I t is hypothesized th a t f e r t i l i t y
levels are determined prim arily by cost fa c to rs, although the opera
tion of other v ariab les, which may also be associated with urbanism-
in d u strial iz a ti on, may mask th e ir e ffe c ts . In other words, i t is
argued th a t urban environments can vary along a number of dimensions
which may lead to high or low f e r t i l i t y . Moreover, while the net
re s u lt of urbanization 1s usually reduction of f e r t i l i t y le v e ls , th is
1s not inevitably the case. The chief hypothesis of the economic
perspective i s , i f other relevant variables associated with urbanism
and in d u stria liz a tio n are held constant, there is a negative
association between cost and f e r t i l i t y .
7
A second perspective im p licit in the work of numerous
students of the urban scene suggests other important dimensions of
urban stru c tu re . In addition to such economic variables as income,
u t i l i t y , or consumer a lte rn a tiv e s , there are variables such as
m igration, education, family stru c tu re , social m obility, e tc . I t is
therefore possible to develop hypotheses about the in te rre la tio n sh ip
of these with f e r t i l i t y , and thereby construct a more general model
re la tin g the c ity to f e r t i l i t y .
Summary and Objectives
This d isse rta tio n is based upon the premise th a t f e r t i l i t y
can meaningfully be studied from an ecological perspective. Within
the context of an "ecosystem" f e r t i l i t y is the dependent v ariab le,
determined by various stru ctu ra l or organizational properties of
c i tie s .
The basic analytical goal is insight into those factors which
determine f e r t i l i t y levels in c i tie s . This implies developing a
model which specifies some of the dimensions along which urban
environments may vary, and linking these to f e r t i l i t y . Analysis is
concerned prim arily with United States c i tie s in 1960. Sources of
hypotheses are urban and economic theory. Inductive comparisons are
made between the determinants of to tal and black f e r t i l i t y .
CHAPTER II
ECO N O M IC A N D SOCIAL DETERM INANTS
OF FERTILITY
In th is chapter, economic and sociological lite r a tu r e relevant
to stru ctu ra l determinants of f e r t i l i t y is reviewed. The review
r e fle c ts three d is tin c t perspectives: economic, sociological, and a
combination of these two. Becker (1960) il lu s tr a t e s the purely
economic perspective. He maintains th a t " . . . although no single
variable in the Indianapolis survey explained more than a small
fractio n of the v ariation in f e r t i l i t y , economic variables did b e tte r
than most." He consequently proposed a purely economic approach to
the analysis of f e r t i l i t y .
Other economists and sociologists have taken a d iffe re n t
positio n . Duesenberry (1960), in a reply to Becker, contends th a t
perhaps in no other area is economic theory less applicable given the
importance of normative constraints arisin g out of membership in such
categories as c la ss , religious groups, or race. Sim ilarly, Bruton
(1965:268) contends th a t formal economic theory is inappropriate.
He remarks th a t " . . . the freedom permitted the parents on family
size decisions is so constrained by in s titu tio n a l factors th a t
applying any s o rt of formal [economic] decision model is not lik ely
to help much."
8
9
A view which is emerging in the work of many economists is th at
these models are fa r from mutually exclusive. In f a c t, the re la tiv e ly
poor success of e ith e r in independently explaining f e r t i l i t y suggests
a need fo r th e ir inclusion in a single perspective. Furthermore, i t
is often d if f i c u lt to c le a rly place p a rtic u la r variables into one
class or the other—for example, is labor force p articip a tio n of
women to be considered an "economic" or a "sociological" variable?
Economic Determinants
Macroeconomic theory is p a rtic u la rly compatible with human
ecology and is emphasized here because i t is customarily defined as
"analysis employing aggregates.” That i s , the principal concern of
macroeconomics is explanation of the behavior of aggregated phenomena,
such as the price and employment levels of to ta l economies. (See, for
example, S irkin, 1961.) Such an approach can readily be adapted to an
application in urban environments.
Heer (1966) maintains th a t economists have developed two
competitive models re la tin g f e r t i l i t y to economic variables. One
approach tre a ts economic development as an inhibiting facto r in
f e r t i l i t y . This approach is best exemplified in tra n s itio n theory and
is supported by the fa c t th a t presently developed countries have
reduced th e ir f e r t i l i t y markedly below preindustrial le v e ls. Further
supporting evidence is the inverse re la tio n between income and
f e r t i l i t y in these countries. The Malthusian tra d itio n is the
opposing perspective, in which i t is maintained th a t economic develop
ment raises f e r t i l i t y because higher incomes permit populations to
10
support la rg e r fam ilies.
An example of the view th a t there is an inverse rela tio n
between economic change and f e r t i l i t y is the work of Okun (1958). H e
attempts to explain family size without making assumptions about
in s titu tio n a l fa c to rs , which are taken as exogenous. Populations
choose the level a t which they w ill "consume" children and other goods
in accordance with rational consideration of the combination of these
two which w ill y ield the g reatest s a tisfa c tio n .
Okun makes a d istin ctio n between children and other commodi
ti e s . The cost of ch ild ren , or the expenditures on them, are a
function of parental s ta tu s . That i s , a child must be raised in
accord with parents' sty le of l i f e , and i t is therefore probable th a t
th e ir cost will ris e more rapidly than expenditures on other
commodities. This is a key assumption of Okun's approach and leads
to the hypothesis th a t aggregate changes in income w ill be inversely
rela ted with changes in f e r t i l i t y . Okun maintains th a t the decline
in Western f e r t i l i t y is explained by th is hypothesis.
There are problems with th is explanation. Most important is
th a t the assertion th a t f e r t i l i t y w ill be related to income is greatly
oversim plified. Other variables which might also be important are
not specified. For example, what i f risin g income is accompanied by
technological changes which reduce the cost of children? Will the
inverse re la tio n s t i l l hold?
This d iffic u lty is partly overcome by Becker (1960), who
illu s tr a te s the perspective th a t economic development is positively
11
correlated with f e r t i l i t y . Becker's scheme, emphasized here, is
becoming increasingly dominant in economic theory, as will be seen
below.
Although th e ir approaches are generally sim ilar, Becker makes
certain d istin ctio n s not made by Okun. Like Okun, Becker maintains
th a t children can be considered consumer goods from which s a tisfa c tio n s
are derived. However, unlike Okun, he in s is ts th a t children are no
d iffe re n t from other consumer commodities. He points out th a t
economists have shown th a t at higher income levels there is a tendency
fo r both more goods and higher quality goods to be purchased. However,
quality e la s t ic i tie s tend to be larg er than quantity e la s tic itie s * —
which means th a t a t higher income levels more emphasis will be placed
upon quality than on quantity. An illu s tr a t iv e analogy is found in
fam ilies, where a high income unit will probably have more cars than
a low income u n it, but will certain ly have b e tte r cars; o r, a higher
income u n it may only have one house, lik e i t s lower income counterpart,
but i t will be of much higher q u ality .
Becker hypothesizes th a t an increase in income will increase
both the quantity and quality of children, with the former increase
small and the l a t t e r increase large. Consequently, there will be a
tendency for higher income populations to have somewhat more children
of a much higher q uality than low income populations.
♦ "E lasticity " is an economic concept which is closely equiva
le n t in meaning to a standardized regression c o e ffic ie n t, and refers
to the amount of change produced in a dependent variable by a unit
change in an independent variable.
12
Becker offers some support for th is hypothesis from the
Indianapolis study. W hen only those couples who successfully planned
both the number and spacing of th e ir children were considered, there
was a positive re la tio n between income and number of children. The
same re la tio n was observed in Stockholm, where i t is argued th a t
contraceptive knowledge has diffused to a ll segments of the population.
Becker attempted to explain the anomaly of the long term
decline in United States f e r t i l i t y (a decline which took place in the
context of risin g income) with reference to other economic factors
which might have acted to counter the effects of income. In p articu
la r , he suggested th a t there may have been a decline in child
m o rtality , an increase in the extent of contraceptive knowledge, and
increases in the cost of children.
Regarding the la s t po in t, Becker suggested th a t the tran sfer
of population from rural to urban environments may have raised the
costs of children fo r the population as a whole. In p a rt, this was
due to the more rapid technological improvement in the marketplace
than in the home, thereby raisin g the costs of children since time and
e ffo rt spent on them meant foregoing other opportunities. Also
important were changes in educational levels and decline in discrim ina
tion against women, both of which would tend to lower f e r t i l i t y .
Becker's model is more sophisticated than Okun's since he
allows for the operation of variables other than income. That i s , he
argues th a t increases in income w ill produce higher f e r t i l i t y , i f other
factors do not in te rfe re .
13
Spengler (1959) has proposed e sse n tia lly the same hypothesis.
He argued th a t when other conditions "stay put," age sp ecific f e r t i l i t y
tends to be po sitiv ely correlated with per capita income. However,
other conditions are rarely constant and, in f a c t, usually a c t in the
opposite d irection with more fo rce, and the net re s u lt is th a t per
capita income is ty p ically inversely rela ted to f e r t i l i t y .
Business Cycles and F e r tility
Studies of the relatio n sh ip between business cycles and
f e r t i l i t y o ffer important support for economic hypotheses. A large
number of such studies have shown a positive association between these
two variables. In one such study, Galbraith and Thomas (1941)
obtained a correlation of +.80 between deviations in trends from
employment and f e r t i l i t y for the years 1919 to 1937, with births
lagged one year. They conclude th a t employment operates both d ire c tly
and in d ire c tly upon f e r t i l i t y . A d ire c t e ffe c t is upon number of
b irth s , while an in d ire c t e ffe c t is through the impact of employment
upon marriage ra te s. (Examples of other studies which reached sim ilar
conclusions include: Ogburn and Thomas, 1922; Thomas, 1927; and
Hexter, 1925.)
However, the v a lid ity of co rrelatin g deviations from trends
has been questioned by Kirk (1960). Although he found th a t deviations
from trends in indexes of in d u strial production, personal income, and
employment were highly correlated with f e r t i l i t y , Kirk was dubious
14
about th e ir importance. He commented th a t: ", . . while the
deviations from trends of f e r t i l i t y rates seem to move in the same
d irection as the trend deviations of economic in d ic a to rs, the former
series [are] in many respects quite independent of economic conditions"
(Kirk, 1960:254). He concluded th a t only "surface waves" are
influenced by economic conditions, while the underlying trends in
f e r t i l i t y are re la tiv e ly independent of economic variab les. Economic
factors are best thought o f, Kirk suggested, as "conditioning fa c to rs,"
rath er than as underlying causes. In a reply to Kirk, Thomas (1960)
contended th a t these co rrelatio n s are highly im portant, and represent
the best examples of "sociological laws."
Income and F e r tility
One of the f i r s t to examine the im plications of Becker's
model for countries was Heer and Turner (1965). In a study of Latin
America they observed th a t countries in which f e r t i l i t y was higher thar
would have been anticipated on the basis of various indexes of
economic development had also experienced recent and large increases
in income. Countries in which lower than expected f e r t i l i t y was
observed tended to have experienced unusually low increases in income.
These data led Heer (1966) to t e s t Becker's hypothesis th a t
while the d ire c t e ffe c t of income increase is to ra ise f e r t i l i t y , the
net re s u lt may be a decrease in f e r t i l i t y . This is because an
increase in income usually is also associated with factors whose long
term effects in decreasing f e r t i l i t y may be stronger, unless per
capita income is increasing a t a very rapid ra te .
15
Heer examined th is hypothesis with data from forty-one nations.
He hypothesized th a t i f other variables associated with economic
development are co n tro lled , income is p o sitiv ely rela ted to f e r t i l i t y .
He controlled fo r three variables which were expected to depress
f e r t i l i t y . F ir s t, education was expected to lower f e r t i l i t y because
of the association of th is facto r with knowledge of b irth control
techniques. Second, population density was expected to lower f e r t i l i t y
because of the increased cost of space associated with high density.
F in ally , decreases in in fan t m ortality were expected to lead to lower
f e r t i l i t y because under conditions of low in fan t m ortality i t is no
longer necessary for populations to produce as many children to ensure
a fixed number of survivors. Note th a t with the exception of educa
tio n , Heer's in te rp re ta tio n s of these hypotheses were in terms of
economic premises.
Heer's findings were consistent with his hypothesis. The zero
order co rrelatio n between income and f e r t i l i t y was -.4 5 , while the
p a rtia l co rrelatio n with education, in fan t m o rta lity , and density held
constant, was +.10—a small but consistent re la tio n . Also as
a n tic ip a te d , the p a rtia l co rrelatio n s between f e r t i l i t y and education,
in fan t m o rtality , and density were moderate but in the anticipated
d ire c tio n — -.2 4 , +.42, and -.4 0 , respectively.
A number of e a r lie r studies also provide data consistent with
Becker's model. Weintraub (1962), using data from th irty developed
and underdeveloped co u n tries, computed p a rtia l co rrelatio n s between
f e r t i l i t y and each of the following: income, per cent employed in
a g ric u ltu re , and in fan t m o rtality . The p a rtia l between income and
16
f e r t i l i t y was +.25, which would have been predicted from Becker's
model. Also consistent was the p a rtia l co rrelatio n between in fan t
m ortality and f e r t i l i t y , which was a substantial +.78.
A sim ilar study by Adelman. (1963) is suggestive, although her
fa ilu re to rep o rt e ith e r p a rtia l correlations or standardized regres
sion co efficien ts makes her data somewhat d if f i c u lt to evaluate. Her
independent variables are per capita income, per cent population
employed outside of ag ric u ltu re , and population density. The dependent
variable is female age sp ecific b irth rates in seven fiv e year age
categories.
Adelman's data were consistent with her hypotheses. Thus,
the sign of the regression co efficien ts for education, per cent
nonagricultural employment, and density, were negative, while th a t fo r
income was p o sitiv e. Although the regression co efficien ts could not
be d ire c tly in terp rete d , the consistency of th e ir sign across a ll age
le v e ls, as well as th e ir general tendency to be s t a t i s t i c a l l y s ig n if i
cant, lends additional support to her hypotheses.
A study by E asterlin (1965) shows the importance of changing
patterns of demand for labor, a variable which is closely rela ted to
income. He analyzed the re la tiv e em ployability of young persons 20-29
and found th a t during the postwar period (through the 1950's), th is
group experienced high income re la tiv e to older groups. This re la tiv e
advantage was attrib u te d to two fa c to rs: (1) small size of th is
group due to the low f e r t i l i t y of the 1930's, in a period when the
general demand fo r labor was high; and (2) re la tiv e educational
17
advantage. These factors acted to sustain high f e r t i l i t y into the
f i f t i e s . A dditionally, Easter!in (1966) suggested in a more recent
paper (1966) th a t recent declines in f e r t i l i t y may be a ttrib u te d to a
lessening of the re la tiv e advantages of th is group.
Gendell (1967) showed th a t economic development did not
appreciably a ffe c t f e r t i l i t y in Brazil between 1920 and 1960. Brazil
was one of the few Latin American countries to experience a substantial
growth in real per capita income during th is period and her in d u strial
stru ctu re moved stead ily away from a primary base toward a secondary-
te r tia r y base, and il lit e r a c y declined ste a d ily . The fa c t th a t
f e r t i l i t y continued high in the face of th is development is consistent
with Becker's model; th a t i s , i t appears th a t the e ffe c t of changes
in occupational and educational s tru c tu re , which should have reduced
f e r t i l i t y , were balanced by increases in income. However, since
Gendell made only lim ited use of correlation an aly sis, and no use of
p a rtia l correlation or standardized regression c o e ffic ie n ts, the
im plications of his findings are uncertain.
One of the most thorough examinations of the re la tio n of
economic and social variables to f e r t i l i t y was made by Friedlander
and S ilver (1967). Their analytic strategy involved computation of
separate regression equations for each of three separate levels of
development. To reduce the e ffects of multi col lin e a r ity , the number
of independent variables included in each regression equation was
usually fiv e . Four variables—measures of income, education, child
m o rtality , and population density—were included in each equation, and
a series of other variables of less th eo retical in te re s t a lte rn a te ly
18
substituted as the f i f t h . However, since each independent variable
was ty p ically measured with a number of d iffe re n t indexes—and these
were varied throughout the analysis—the re s u lt was a large number (42)
of separate regression analyses.
Independent variables in these equations were evaluated in
terms of the consistency with which a given co e ffic ie n t reached
s ta t is tic a l significance and whose sign was in the predicted d irection
in the series of equations in which i t appeared. Of importance for
the present study are th e ir data on income, education, density, social
m obility, and sta tu s of women. Regarding income, they found th a t
various indexes tended to have positive signs in developed countries,
negative in underdeveloped. A number of economic in terp reta tio n s were
suggested; for example, i t is possible th a t in underdeveloped
countries quality e l a s t ic i tie s may be larg er than quantity e l a s t ic
i t i e s , as Okun proposed.
There was a consistent tendency fo r both measures of education
to have negative regressions on f e r t i l i t y . This was interpreted as
re fle c tin g a number of fa c to rs: achievement motivation, contraceptive
knowledge, change of ta stes away from children toward other sources of
s a tis fa c tio n , and differences in domestic versus market productivity.
They also found th a t the statu s of women tended to be
positively related to f e r t i l i t y . Such a finding is inconsistent with
current theory, but th e ir operational d efin itio n of female statu s as
the ra tio of male to female il lit e r a c y rates may r e f le c t only one
meaning of th is complex concept.
19
The coefficients involving density were usually negative and
s ta t is tic a lly sig n ific a n t, despite the fa c t th a t i t was defined as the
ra tio of to tal population to to tal land area, rath er than population
to usable land area. They interpreted th is relationship as an
indication of the re la tiv e ly high cost of living space when popula
tions are heavily s e ttle d .
F inally, Friedlander and Silver argued th a t urbanization is
inversely related to f e r t i l i t y because of the higher costs of raising
children in urban as opposed to rural areas. However, they base th is
conclusion upon an unusual in terp reta tio n of th e ir data. That is ,
they found th a t the regression of per cent non-agricultural population
owning i t s own business was considerably higher than the coefficients
for extent of ag ricu ltu ral population. Such data suggest, they
maintained, th a t even in urban environments couples will have large
fam ilies when i t is economically advantageous.
Friedlander and Silver concluded th a t th e ir data substantiate
the usefulness of a "demand" approach in the analysis of f e r t i l i t y .
However, th e ir reliance upon unstandardized regression coefficients
and the determination of the importance of variables in terms of
consistency of sign and s ta t is tic a l significance makes th e ir data
d if f ic u lt to in te rp re t. Had they u tiliz e d p a rtia l correlation or
standardized regression c o e ffic ie n ts, th e ir analysis would have been
much clearer. Furthermore, th e ir in terp retatio n s are frequently open
to question because of the imprecision of th e ir indexes and th e ir
tendency to make inferences about the behavior of individuals based
upon correlations of aggregated t r a i t s .
20
Summary
Data reviewed in th is section are consistent with the
hypothesis th a t some economic variables are important determinants of
f e r t i l i t y . That i s , populations seem to respond to cost factors such
as density and increase th e ir production of children a t higher income
le v e ls , other things being equal. Furthermore, the inverse co rrela
tion with in fan t m ortality suggests th a t populations ratio n a lly adjust
numbers of b irth s to levels necessary to maintain a constant family
size.
Social Determinants
In th is sectio n , some of the "social" dimensions associated
with urban stru ctu res are considered. These variables have rarely
been applied to aggregated d ata; ra th e r, the tendency has been to
examine th e ir re la tio n with f e r t i l i t y a t the individual lev el. Con
sequently, the discussion only suggests hypotheses which might be
appropriate to aggregated data. Five broad classes of variables
are reviewed: labor force p articip a tio n of women, education,
occupation, and horizontal and v e rtic a l m obility.
Race is another important variable because black-white d if
ferences in f e r t i l i t y have been marked. However, i t makes l i t t l e
sense to conceive of th is fa c t as associated with urban stru ctu re s.
There is reason to believe th a t the f e r t i l i t y of racial groups is
rela ted in d iffe re n t ways to the stru ctu ra l variables considered. I t
is probable th a t black f e r t i l i t y is affected by variables which could
not meaningfully be applied to white f e r t i l i t y , such as segregation
21
or family disorganization. Consequently, race is introduced for
comparative purposes. The same hypotheses w ill be tested for both
blacks and whites where available data permit.
Labor Force P articip atio n of W om en
As with income, most of the work on the re la tio n between labor
force p articip a tio n of women and f e r t i l i t y has been a t the individual
le v el. The re su lts are uniform in support of the conclusion th a t
there is an inverse re la tio n sh ip , leading Blake (1965:1195) to s ta te
th a t, "Female labor force p articip a tio n has long been known to bear
one of the most impressive relationships to family size of any
v aria b le." In fa c t, Blake found th a t fo r a sample of United States
high school and college students, even desire to work was inversely
correlated with desired family siz e . This relatio n sh ip has been
interpreted as re fle c tin g e ith e r role c o n flic t fo r women who must
choose between a career or a family, or the fa c t th a t small fam ilies
present women with the opportunity for outside employment (cf. Tien,
1967:222). These possible in terp reta tio n s show th a t the d irection of
the causal rela tio n sh ip is unclear.
Regarding individual d ata, United Nations experts (1953:88-89)
summarize data for a number of censuses in several countries which
show th a t married women who are employed have fewer children than do
other married women. P articu larly s ig n ific a n t are data from the
Swedish Census of 1935-36, which show th a t married women who had been
gainfully employed throughout th e ir married lives had fa r fewer
children than did women who had never been employed. W omen with
22
employment patterns between these two extremes also had interm ediate
levels of f e r t i l i t y . However, the United Nations experts point out
th a t these data do not indicate whether employment leads to smaller
fam ilies, or infecund women are selected into the labor force. In
f a c t, Freedman, Whelpton, and Campbell (1959) suggest th a t th is la s t
p o s sib ility accounts for p art of the rela tio n sh ip between f e r t i l i t y
and employment of women.
Although the existence of th is relatio n sh ip a t the individual
level is well documented for developed co u n tries, data fo r under
developed countries are inconsistent. Thus, Gendell (1965) reports
th a t in the tra d itio n a l sector of Indian so ciety , employment of women
seems to have l i t t l e or no influence upon f e r t i l i t y . Similar findings
are reported fo r Turkey by Stycos and Weller (1967). These authors
interviewed 2,700 women in three size classes of community, and found
th a t the re la tio n between female employment and f e r t i l i t y v irtu a lly
disappeared when any one of a number of controls were introduced.
Control variables included: education, rural-urban residence, and
length of marriage.
In co n tra st, survey data from a number of other underdeveloped
countries indicate th a t f e r t i l i t y and employment of females are
inversely re la te d . For example, Miro and Rath (1965) drew representa
tiv e samples of 2,000 women aged 20-50 from each of the following
c i t i e s : Rio de Jan eiro , Panama, and San Jose. Non-employed females
averaged 3.5 , 2 .5 , and 3.7 liv e b irth s , respectively. Working women
in these same c itie s averaged 3.0, 1 .8 , and 1.9 live b irth s.
23
However, co rrelatio n s a t the individual level t e l l us l i t t l e
about the impact of labor force p a rtic ip a tio n of women a t the com
munity le v el. That i s , even i f the impact of th is variable is
substantial fo r in d iv id u als, i t may be negligible for communities.
Furthermore, a t the aggregate level the in te rp re ta tio n of the
rela tio n sh ip is very d iffe re n t. Rather than in te rp re t in terms of
such intervening variables as "role co n flict" i t is argued th a t
employment levels of women r e f le c t variatio n in th e ir integration into
the broader society. (This point w ill be taken up in more d etail in
Chapter I I I .)
Collver (1968) conducted an important study of the aggregate
re la tio n between labor force p articip a tio n of women and f e r t i l i t y in
United States m etropolitan areas. He found th a t, in general,
communities with high labor force p a rtic ip a tio n of women tended also
to have low m arital f e r t i l i t y and low proportions married. However,
the pattern of co rrelatio n s was d iffe re n t fo r young as opposed to
older age groups. For the younger group, aged 20-24, there was a very
high negative co rrelatio n between labor force p articip a tio n of women
and per cent married (-.80) and a very small positive correlation
(+.08) between labor force p artic ip a tio n and number of children ever
born to women in th is age category; Collver in te rp re ts these
co rrelatio n s as indicating th a t the e ffe c t of labor force p a rtic ip a
tion was in d ire c t, operating through i t s effects upon marriage. For
older women, aged 30-34, the co rrelatio n between labor force
p articip a tio n and f e r t i l i t y was moderate and negative (-.3 7 ), while
th a t between labor force p a rtic ip a tio n and per cent married was
24
negligible (-.0 1 ), suggesting th a t labor force p articip a tio n affe c ts
f e r t i l i t y d ire c tly for th is age category.
An e a r lie r study of f e r t i l i t y in 284 United States c itie s with
1920 census population of 25,000 or more by Thompson (1931) showed
th a t c itie s with high proportions of women in the labor force tended
to have low f e r t i l i t y . Lorimer and Osborn (1934), using Thompson's
data, found th a t the correlation between these two variables was -.38.
Sim ilarly, Hashmi (1964) found an inverse relatio n sh ip between
f e r t i l i t y and labor force p articip atio n of women when the u n it of
measurement was Chicago Census tra c ts in 1950.
Studies of other countries using aggregated data are consistent
with these findings. Collver and Langlois (1962) found th a t labor
force p articip a tio n rate s for women and f e r t i l i t y were inversely
correlated in the metropolitan areas of twelve countries. Correlations
ranged in magnitude from -.12 in India to -.64 in Canada. Using
aggregated data fo r e n tire countries, these authors found a correlation
of -.6 0 . Heer (1964) observed the same pattern in a study of
departments in B olivia, Ecuador, and Peru.
In conclusion, employment of females tends to be inversely
related to f e r t i l i t y . This re la tio n is especially clear for
aggregated d ata, although the pattern fo r individual data is not so
unambiguous. However, fo r the United S tates, both kinds of data are
consistent and indicate an inverse relatio n sh ip .
Education
One of the most p e rsiste n t co rrelates of f e r t i l i t y is
25
education. A large number of studies have shown these two to be
related in a v ariety of developmental and cultural s e ttin g s . Further
more, there is a theoretical reason to suppose th a t level of education
is rela ted to a se rie s of other stru ctu ra l v ariab les, such as division
of labor and m obility. Consequently, education is analyzed to
evaluate both i t s d ire c t and in d ire c t e ffe c ts.
As fa r back as data are av ailab le, individual correlations
consistently show an inverse re la tio n . United States Census data
analyzed by G rab ill, K iser, and Whelpton (1958) show th a t, in 1950,
married women with four years of college or more had a completed
family size of 1.4, compared with 2.6 for women with an elementary
school education or le ss. The same pattern was observed fo r married
rural farm women, with completed family size ranging from 1.8 for
college educated women, to 5.1 for women with elementary or no
schooling. Duncan (1965) showed th a t w ithin a ll occupational cate
gories of husband, as education of wife increased, f e r t i l i t y
decreased. (For other United States studies see, fo r example, Kiser,
1942; and Whelpton and Kiser, 1943.)
Analysis of other countries reveal the same p attern . For
example, an early study of d iffe re n tia l f e r t i l i t y in Sweden by Edin
and Hutchinson (1935) showed th a t th is d iffe re n tia l existed ju s t a fte r
World War I. A la te r study of the same country indicated th a t the
same relatio n sh ip between these two variables existed in the post
World War II period (Moberg, 1950).
At the aggregate le v e l, p arallel co rrelatio n s have been
26
reported. For the United S ta te s, Hashmi (1964)showed th a t as the
educational level of Census tra c ts in Chicago increased, f e r t i l i t y
decreased. Bogue (1969) correlated the point a t which a given country
was at in making the demographic tra n s itio n with a series of other
v aria b les, including education and urbanization, measured by per cent
urban. His measures of education co nsistently correlated more
highly with th is index than did other v a ria b le s, including urbaniza
tio n . Thus, per cent i l l i t e r a t e correlated -.7 7 , per cent educated
a t "level one" or higher +.75, and per cent in school +.79.
Urbanization, on the other hand, correlated +.55. Furthermore, when
education a t "level one" was co n tro lled , the correlation with other
variables and f e r t i l i t y was e ith e r greatly reduced or disappeared.
These data led Bogue to conclude th a t: "A major driving force behind
f e r t i l i t y control appears to be education" (1969:676).
A possible e ffe c t of education upon f e r t i l i t y is its impact
upon knowledge of b irth control and w illingness to implement i t . Not
only are b e tte r educated individuals more lik e ly to plan th e ir
fam ilies, educational levels are rela ted to the general flow of
contraceptive knowledge in the population. Heer (1966) examined th is
hypothesis and suggested th a t an increase in education is lik e ly to
lead to increase in the " . . . flow of communications of a ll types
. . . [and] an increase in communications sp ec ific a lly concerned with
the technology and consequences o f b irth control practices" (1966:428).
Heer operationalized education in terms of newspaper circ u latio n per
thousand and found th a t the co rrelatio n between th is variable and
f e r t i l i t y , with such variables as income con tro lled , was -.24.
27
Considering only the education of females, a number of
studies have shown th a t the most sig n ifican t variable in u tiliz a tio n
of contraception is w ife's education. For example, Driver (1963)
found th a t among samples of Indian, women the c le a re s t co rrelate of
low f e r t i l i t y was education of the wife.
Finally, Stycos (1968) correlated a series of measures of
education with the child-women ra tio and the crude b irth rate in a
number of Latin American countries. Correlations ranged from -.18
to -.62. These data tend to support the hypothesis th a t education
is an important facto r in f e r t i l i t y .
Occupation and F e r tility
A substantial amount of data is available which indicate th a t
f e r t i l i t y varies among occupational groups, a t le a s t in developed
countries. However, whether or not these differences e x is t independent
of the effects of income, education, and age a t marriage is open to
question. Furthermore, almost a ll analyses have been a t the individual
le v e l, and th e ir in terp reta tio n s for aggregates is unclear.
Bogue (1969) showed th a t, fo r the United States in 1960, there
were moderate differences among major occupational groups in th e ir
f e r t i l i t y . For example, the lower-white c o lla r group had the lowest
f e r t i l i t y (measured by children ever born per woman). The most
f e r t i l e was the blue co lla r group, followed by the upper-white co lla r
group. However, most of these differences disappeared when education,
income, and age a t marriage were controlled, with the exception of
farmers and laborers.
28
Apparently the same patterns e x is t in other developed
countries. United Nations (1953:88), commenting on studies conducted
in a number of countries during the 1930's concluded th a t: "The
lowest f e r t i l i t y was no longer found among 'professional workers,'
but rath er among certain other categories of the 'white c o lla r' group
generally believed to have a lower income and 'so cial s ta t u s '—e . g . ,
c le ric a l workers, o f fic ia ls in public adm inistration, owners of
businesses, e tc ."
Data based upon the 1954 French census support th is conclusion
(Febvay, 1959). These data indicate th a t wives of men employed in
trade and offices had fewer children than e ith e r wives of laborers
or those employed in the "lib eral professions."
These occupational differences are d iffe re n t from those
observed during early phases of the demographic tra n s itio n . Numerous
studies of th is pattern show a c lear inverse association when
occupations are arranged by s ta tu s . The most common in terp reta tio n
is th a t the higher sta tu s occupational groups were the f i r s t to acquire
the rational orientation described by Weber and therefore were the
f i r s t to perceive the economic advantages of small fam ilies. These
a ttitu d e s diffused successfully to lower statu s occupational groups,
ultim ately producing a clear inverse re la tio n between statu s and
f e r t i l i t y (cf. Banks, 1954; Bashers, 1967). There is less consensus
upon explanations of current p attern s. One in te rp re ta tio n is th a t
urban-industrial countries are in the midst of a tra n sitio n to a
positive re la tio n between income-occupationand f e r t i l i t y . -Eventually,
i t is argued, fam ilies will have as many children as they can afford,
29
based upon rational calculations of co st. The present pattern
presumably represents a tra n sitio n a l phase (cf. Hawley, 1950;
Petersen, 1961).
Data fo r underdeveloped countries re la tin g occupation to
f e r t i l i t y are scarce and the re su lts contradictory. Stycos (1968)
drew a sample of about 2,000 currently married women from the popula
tion of Lima, Peru. Interviewers categorized respondents according
to social c la s s , a principal determinant of which was occupation.
There was a positive re la tio n between preferred number of children and
c la ss , but a negative re la tio n between these two variables and actual
number of children. Stycos interpreted the positive re la tio n in
economic terms; he showed th a t lower class respondents are aware of
the deletorious economic effects of additional children. However,
they are unable to implement th e ir desire for smaller fam ilies for a
number of reasons, including lack of information about b irth control.
In conclusion, data on developed countries indicates th a t
occupation is related to f e r t i l i t y , with lower-white c o lla r workers
having the lowest le v els. Although the diffusion hypothesis often
offered as an explanation is usually cast in individual terms, i t w ill
be argued in Chapter I I I th a t an aggregate in te rp re ta tio n of this
hypothesis is more appropriate.
Social Mobility and F e r tility
A popular hypothesis is th a t social m obility is causally
rela ted to f e r t i l i t y . An early statement of th is so-called "mobility
hypothesis" was by Dumon (referenced in Freedman, 1961 -1962':40-)oand
30
maintained th a t: "ju st as a column of liquid must be thin in order
to ris e under the force of c a p illa r ity , so must a family be small to
ris e in the social order." Explanations of th is rela tio n have
invariably been a t the individual level and usually contain one or
both of the following elements: (1) couples must choose between large
fam ilies and upward m obility, since devotion of time and energy to one
may mean th a t re la tiv e ly l i t t l e is available for the other; or (2)
social mobility is associated with secularization and individuation
with consequent weakening of trad itio n al norms regarding f e r t i l i t y .
Numerous studies in a variety of cultural settin g s have given some
support to these hypotheses (cf. Freedman, 1961-1962).
A recent d efin itiv e study of the re la tio n between social
m obility and f e r t i l i t y was made by Blau and Duncan (1964). They found
th a t while mobility d e fin ite ly has an e ffe c t, i t is small re la tiv e
to such factors as occupation and education. Generally, upwardly
mobile couples had fam ilies which were intermediate in size between
class of origin and class of destin atio n . A social mobility hypothesis
was also examined in the cla ssic Indianapolis study (Whelpton and
Kiser, 1958). D istinctions were made between planned family size and
success in family planning. They found th a t so cially mobile couples
plan smaller fam ilies than do non-mobile couples. However, they were
in-between class of origin and class of destination in th e ir planning
effectiveness.
Regarding intergenerational mobility and f e r t i l i t y , the data
are even more sketchy. One study, an unpublished doctoral d issertatio n
by Yell in (1955) showed th a t upward m obility had no influence upon
31
number of children. Yell in studied urban salesmen, engineers, and
bankers.
Friedlander and S ilver (1967) attempted to evaluate the effects
of social m obility upon societies.. Although they concluded th a t social
m obility is in fa c t related inversely to f e r t i l i t y a t th is le v e l, the
data presented in support of th is conclusion are not convincing. Not
only was the regression co e ffic ie n t in the predicted d irection in only
one se t of countries, they published only unstandardized co efficien ts
rath er than p a rtia l correlations or betas. Consequently, there is no
way of estim ating the re la tiv e importance of th is variable. F in ally ,
th e ir index of social mobility was highly s p e c ific , operationalized
as the geometric ra te of growth of the gross domestic product.
Horizontal Mobility
Turning to the rela tio n sh ip between resid en tial mobility and
f e r t i l i t y , most research has concentrated upon suburbanization.
I t has often been hypothesized th a t couples who move from the c ity to
suburbs are motivated by a desire fo r large fam ilies, a value which
can presumably be maximized in suburban environments. For example,
Bell (1958) trichotom izes fam ilies into those which are oriented
prim arily toward: familism, consumption, and m obility. To a large
degree these o rien tatio n s are defined in mutually exclusive terms.
Bell hypothesizes th a t couples who move from the c ity to suburbs are
drawn prim arily from the f i r s t type, while those who remain are drawn
from the la s t two. Although not sp ec ific a lly considered by him,
B e ll's approach is consistent with the hypothesis th a t in-migrants
32
into c itie s will be prim arily oriented toward consumption and/or
social m obility.
Bogue (1969) showed for United States Census data th a t the
f e r t i l i t y ra tio s of migrant women are considerably higher than those
of non-migrant women. However, th is appearance of high f e r t i l i t y was
actually due to the fa c t th a t migrants are concentrated in the ch ild
bearing ages. W hen age standardized rate s were compared, the f e r t i l i t y
of migrants was almost the same as for non-migrants.
In conclusion, there is some reason to believe th a t both social
and horizontal mobility are associated with f e r t i l i t y . However, there
is some question as to the independent e ffe c t of horizontal m obility.
I t has been argued th a t horizontal mobility is largely a refle ctio n
of v ertical m obility. That i s , research has shown th a t long distance
migrants are usually motivated by a desire to improve th e ir economic
sta tu s. For example, an im plication of B ell's theory is th a t migrants
into c itie s are oriented e ith e r toward upward m obility or consumption.
On the other hand, Bogue's data suggest th a t migrants have unusually
high f e r t i l i t y , due prim arily to th e ir age stru ctu re . In Chapter III
an attempt will be made to s o rt out the independent effects of these
fa c to rs .
Black F e r tility
Black birth rates have always been higher than those of
whites. Although data are not detailed fo r the e a r lie s t periods, there
is no question about the existence of a d iffe re n tia l. From about 1900
to 1947, black and white rates converged stead ily . However, in 1947,
33
white f e r t i l i t y began to level o f f , while black f e r t i l i t y continued
to r is e , resulting in divergence (Petersen, 1961). That current
differences are not due to differences in age stru ctu re was shown by
Kiser (1958) who found th a t they existed even when general f e r t i l i t y
rates were compared.
Black f e r t i l i t y seems to re fle c t two major forces which have
affected the in stitu tio n a l stru ctu re of th is population-geographical
and social relocation. As is well known, the black population has
been transformed in a seventy year period from prim arily southern and
rural to prim arily urban and in d u s tria l, a tra n sitio n which took the
United States as a whole more than 150 years.
Geographic relocation took place in two major stages. As the
plantations of the Gulf and southwestern sta te s became major users of
black labor, large numbers moved west (Farley, 1965). Of greater
importance were the great northward migrations which got under way
a t the turn of the century. In 1910, ninety per cent of the nation's
blacks lived in southern s ta te s , almost all of whom were concentrated
in ag ricu ltu re. By 1960, sixty per cent of the black population was
found in the southern s ta te s . Furthermore, in th a t year more than
sixty per cent of the black population was defined as "urban"
(Hamilton, 1964).
The use of percentages masks the great numbers of people
involved in these migrations. For example, in 1960 there were
3,248,736 blacks living in the north and west who had been born in
the south. The peak years of migration were in the 1940's, when more
than one and one-half m illion blacks l e f t the south. Moreover,
34
v irtu a lly all of these migrants s e ttle d in central c i tie s (Hamilton,
1964).
Regarding changes in occupational d is trib u tio n , blacks tended
to move from one low statu s position to another. However, Farley
(1965) maintained th a t a sig n ific a n t portion took higher status
positions. Status considerations asid e, the black population was
obviously required to make massive readjustments to to ta lly d iffe re n t
job and other environmental requirements.
The Tauebers (1965) suggested th a t the character of Negro
migration has changed greatly in recent years. Using 1960 United
States Census data, they showed th a t non-white in-migrants from
non-southern SMSA's tend to be b e tte r educated and more lik ely to have
white c o lla r employment than the indigenous non-white population. In
regard to education, the non-white in-migrants tended to be much lik e
the white population. A very d iffe re n t situ a tio n existed for
southern SMSA's, where in-migrants were ty p ic a lly rural in origin and
below the indigenous white population in educational background.
The possible effe c ts of these factors are outlined in Chapter
II I . The basic hypothesis is th a t high rate s of migration and
occupational change lead to changes in the organization of the black
population, including family disorganization. Hypotheses linking
these variables are explicated in Chapter I I I .
CHA PTER III
TH EO RY A N D HYPOTHESES
In th is chapter a number of sp ecific hypotheses are presented
which re la te f e r t i l i t y to several dimensions of urban stru c tu re .
These dimensions usually are not d ire c tly comparable to the economic
and social factors considered in Chapter II because these factors have
not been studied within an appropriate ecological framework. To avoid
th is problem, some of the possible dimensions along which urban
stru ctu re may vary are specified. The evidence presented in Chapter II
can then be subsumed under these general dimensions. For example,
the degree to which women are integrated into the broader system is
one such dimension; the labor force p articip a tio n of women is a
re fle c tio n of th is dimension.
I t is not possible to specify al_l_ dimensions along which urban
stru ctu res may vary, p artly because the necessary data are not
available and also because the present s ta te of ecological theory
makes an exhaustive formulation impossible. Furthermore, a key
assumption is th a t urban f e r t i l i t y is determined by a number of
fa c to rs, not all of which are considered here. This is lik e ly to
produce< re la tiv e ly low zero order co rre la tio n s, making d if f i c u lt the
in te rp re ta tio n of b iv a riate hypotheses.
35
36
The following dimensions of urban stru ctu re are considered for
the e n tire community: income, c o st, integration of women, education,
bureaucratization, m igration, and social m obility. A theoretical
ratio n ale is presented re la tin g each of these variables to f e r t i l i t y .
Where appropriate, controls are suggested.
The determinants of black f e r t i l i t y are considered separately
in the data analysis and in te rp re ta tio n section of the study.
In terp re tatio n is com parative--the re la tiv e importance of the various
stru ctu ra l dimensions fo r black as compared to to ta l f e r t i l i t y is
analyzed. However, th is analysis is largely inductive since there are
few a p rio ri grounds for an ticip a tin g re la tiv e magnitudes of regression
or co rrelatio n co efficien ts between the two groups. Consequently,
unless otherwise s ta te d , i t is anticipated th a t hypotheses are relevant
fo r both black and to tal f e r t i l i t y . An exception, for which an
hypothesis sp ecific to blacks is presented, is segregation.
Structural Variables and Hypotheses
Income and F e r tility
Becker (1960) suggests th a t changes in income w ill be posi
tiv e ly associated with changes in f e r t i l i t y . I t is argued th a t a t
higher income levels populations increase th e ir consumption of goods,
and ra is e th e ir production of children. That i s , as income increases,
populations are able to afford more children as well as consumer
goods. Of course, children are not consumer goods in quite the same
sense as cars or re frig e ra to rs since expenditures on the former are
tie d to the population's standard of liv in g . Consequently, a t higher
37
income levels populations are c u ltu ra lly constrained to expend more
on children than they are a t lower le v e ls ; they cannot make the same
choices between q u ality and quantity of children in the same way th a t
they can choose to purchase a Volkswagon and a Ford rath er than a
C adillac. In f a c t , people a t higher income levels are lik e ly to
expend proportionately more on th e ir children since provision of high
qu ality educational and other advantages w ill probably be more
important than a t lower le v els.
In economic terms, i t is hypothesized th a t the quality
e la s tic ity will be larg er than the quantity e l a s t ic i ty , but both are
hypothesized to be p o sitiv e. (" E la s tic ity " is equivalent to a
standardized regression c o e ffic ie n t, and refers to the amount of
change produced in a dependent variable by a u n it change in an
independent v a ria b le .) Therefore, as income increases, populations
are inclined to produce more ch ild ren , but they w ill also d iv e rt a
larg er proportion of th e ir income to increasing th e ir quality .
Consequently, i t is not anticipated th a t u n it changes in income levels
will produce large changes in f e r t i l i t y .
Becker contends th a t the negative re la tio n between income and
f e r t i l i t y usually observed is the re s u lt of other variables masking
the "true" re la tio n sh ip . In other words, the e ffects of variables
which tend to depress f e r t i l i t y a re , in th e ir to ta l impact, stronger
than income. Consequently, an adequate te s t of Becker's hypothesis
requires th a t variables are controlled which might be inversely
associated with f e r t i l i t y but p o sitiv ely related to income.
38
This discussion suggests the following hypotheses:
(la) Income is inversely related to f e r t i l i t y , i f no
variables are controlled.
(lb) Income is positively, related to f e r t i l i t y i f
relevant variables are controlled.
Relevant control variables appear to be education, bureaucra
tiz a tio n , and integration of women. Education is expected to be
po sitiv ely associated with income, but inversely associated with
f e r t i l i t y . Sim ilarly, income is lik e ly to be positively associated
with the extent to which women are integrated into the system, as well
as with bureaucratization, variables which are expected to be inversely
associated with f e r t i l i t y . (Explanations for these hypotheses and
d efin itio n s of concepts will be presented below.)
Cost and F e r tility
The economic model suggests th a t as the cost of children
rise s re la tiv e to consumer goods, these consumer goods are preferred.
Consequently, as the cost of children r is e s , they are lik ely to be
produced in sm aller numbers. At the aggregate le v e l, the cost of
children probably has a number of dimensions, one of which is density.
Under the condition of high density, l i t t l e space is available and
consequently becomes more expensive. Since children require space,
the following hypothesis is suggested:
(2) Density is inversely related to f e r t i l i t y .
39
Integration of W omen and F e r tility
I t has often been hypothesized th a t a substantial portion of
the decline in Western f e r t i l i t y is a ttrib u ta b le to the increased
integration of women into the broader society. Integration of women
refers to th e ir emancipation from strong tie s to the family and th e ir
increased p articip a tio n in other ro les. That i s , integration denotes
tie s to a wide range of in s titu tio n s and a c tiv itie s other than the
family. Since children are an impediment to p articip a tio n in roles
outside the family, i t is anticipated th a t as women become more
integrated into the broader society th e ir f e r t i l i t y w ill f a l l .
This analysis suggests a meaningful in terp reta tio n of the
re la tio n between labor force p articip a tio n of women and f e r t i l i t y .
Labor force p articip a tio n is interpreted as a re fle c tio n of the
degree to which women are integrated into the economy. This leads to
the following hypothesis:
(3) Labor force p articip a tio n of women varies inversely
with f e r t i l i t y .
Education and F e r tility
Education can a ffe c t f e r t i l i t y in several ways. F ir s t, i t
may create in te re s ts and lead to a c tiv itie s incompatible with large
fam ilies. This is true for both males and females, because education
ty p ically o rien ts both toward extra-fam ilial a c tiv itie s . Another
e ffe c t is th a t education leads to increased knowledge and communica
tion about b irth control along with greater w illingness to u tiliz e i t .
Education is also expected to increase bureaucratization and income,
40
both of which are anticipated to be inversely related to f e r t i l i t y .
F inally, migration is expected to be positively associated with
education and f e r t i l i t y . (Explanation of the hypotheses relatin g
bureaucratization and migration to. f e r t i l i t y are presented below.)
This discussion leads to the following hypotheses:
(4a) Education varies inversely with f e r t i l i t y .
(4b) The magnitude of the inverse association between
education and f e r t i l i t y w ill be increased i f migration
is controlled.
(4c) The magnitude of the inverse association between
f e r t i l i t y and education will be reduced i f bureaucrati
zation and income are controlled.
When the effects of education are examined sp ec ific a lly for
females, a fu rth er in terp reta tio n of the integration of women
hypothesis is suggested. As indicated above, as women become more
integrated they no longer confine th e ir a c tiv itie s to the family
(by d e fin itio n ). Moreover, i t is assumed th a t more highly educated
females are more often employed than less educated females.
Therefore, i t is anticipated th a t education w ill operate both
d ire c tly and in d ire c tly upon f e r t i l i t y . To evaluate both types of
e ffe c ts , the integration of women is introduced as a control. I t is
possible th a t the e ffe c t of female education on f e r t i l i t y is accounted
fo r by the e ffe c t of female education on labor force p articip a tio n .
However, th is is not anticipated since higher education of women is
also expected to increase knowledge of b irth co n tro l, w illingness to
41
plan fam ilies, and p articip a tio n in non-economic a c t iv it ie s . Since
migration is expected to be positively correlated with both f e r t i l i t y
and female education, i t w ill tend to suppress the "true" re la tio n sh ip ;
consequently, i t is necessary to control migration when examining the
association between female education and f e r t i l i t y . This discussion
leads to the following hypotheses:
(4d) Education of wamen is inversely rela ted to f e r t i l i t y
i f migration is controlled.
(4e) The magnitude of the inverse association between female
education and f e r t i l i t y will be reduced i f the labor
force p articip a tio n rate of women is controlled.
Bureaucratization
Although few attempts have been made to examine the re la tio n
ship between bureaucratization and f e r t i l i t y a t the ecological le v e l,
there is widespread agreement on the importance of th is variable.
Blau and Duncan (1964) hypothesized th a t bureaucratization is inversely
rela ted to f e r t i l i t y . They maintain th a t bureaucratic structures
fo ste r a G esellschaft ra th e r than a Gemeinschaft system of social
rela tio n sh ip s. In G esellschaft s tru c tu re s, social re la tio n s are
valued prim arily as a means to some other end, not as ends in them
selves. In other words, the goals of G esellschaft systems require
in teractio n in terms of sp ecific role segments and consequently
relationships are re la tiv e ly superficial and other persons are defined
instrum entally.
These orientations pervade all aspects of l i f e , extending even
42
to family s e ttin g s . Blau and Duncan say th a t, "The w illingness to
postpone marriage is a p a rtic u la rly straightforw ard expression of the
w illingness to permit rational d elib eratio n to influence decisions
about the most intim ate social rela tio n s" (1964:416). Such o rie n ta
tions also find expression, according to these authors, in the
rational planning of family siz e .
Blau and Duncan fu rth e r argue th a t white c o lla r workers have
a more d is tin c t G esellschaft o rien tatio n than do blue c o lla r workers.
This is because the former deal prim arily with people and achieve
th e ir goals by influencing others. Blue c o lla r workers, on the other
hand, work with things ra th e r than people and therefore lack the
instrumental o rien tatio n s to others c h a ra c te ristic of the white c o lla r
workers. The authors explain the higher f e r t i l i t y of blue c o lla r
workers in terms of th is more spontaneous a ttitu d e which presumably
extends to decisions about family size.
I t is argued th a t the extent to which communities are dominated
by bureaucratic stru ctu res is negatively associated with f e r t i l i t y .
The more pervasive is the influence of th is organizational p attern ,
the g reater w ill be the diffusion of the bureaucratic ethos. Such
diffusion w ill have a depressing e ffe c t upon f e r t i l i t y . This is the
d ire c t e ffe c t of bureaucratization.
However, the to ta l e ffe c t of bureaucratization is more
complex. Bureaucratization is expected to be causally rela ted to
m igration, education, and income le v e ls. The TauebersJ data indicate
th a t migration streams into and out of large c i tie s are prim arily
43
high income white c o lla r workers. I t is probable th a t bureaucratic
systems require a constant interchange of these kinds of workers.
Consequently, bureaucratization is expected to be positively
associated with migration. However, migrants tend to be concentrated
in the child bearing ages, and consequently, migration is expected
to be positively rela ted to f e r t i l i t y . This suggests th a t the inverse
re la tio n between bureaucratization and f e r t i l i t y is obscured by i t s
effe c ts upon m igration. This leads to the following hypothesis:
(5a) The magnitude of the inverse association between
bureaucratization and f e r t i l i t y w ill be increased i f
migration is controlled.
Also important is bureaucratization's e ffe c t upon income. I t
is lik e ly th a t bureaucratization will be positively associated with
income. Because income is expected to be inversely related to
f e r t i l i t y , i t s e ffe c t is to increase the association between
bureaucratization and f e r t i l i t y . That i s , an in d irect e ffe c t of
bureaucratization is through i t s e ffe c t upon income.
As was suggested above, bureaucratization is expected to lead
to higher educational le v els. Bureaucratic systems are predicated
upon re la tiv e ly high degrees of ex p ertise, and bureaucratic jobs
consequently require more education. Consequently, bureaucratization
is expected to be positively associated with education. Therefore, a
second in d ire c t e ffe c t of bureaucratization upon f e r t i l i t y is through
education. This discussion suggests the following hypothesis:
(5b) If education and income are controlled, the association
44
between bureaucratization and f e r t i l i t y w ill become
less negative.
The relationships discussed in th is section are outlined in
Figure 1. This suggests th a t the d ire c t effects of bureaucratization
upon f e r t i l i t y can be evaluated with the p artial correlation between
these two variables with education, income, and migration controlled.
I t is expected th a t th is correlation will be negative.
Migration
Education
F e r ti1i ty
Income
Bureaucratization
Figure 1.
The following hypothesis is indicated:
(5c) Bureaucratization w ill be inversely related to f e r t i l i t y
i f education, income, and migration are controlled.
45
Migration
As indicated above, migrants tend to be concentrated in the
child bearing ages and consequently are expected to have high f e r t i l
ity . However, they are also lik e ly to be higher in socio-economic
c h a ra c te ristic s than non-migrants. These facts suggest th a t the
re la tio n between migration and f e r t i l i t y is multidimensional. To the
extent th a t a more bureaucratic stru ctu re is re fle c te d , lower
f e r t i l i t y is an ticip ated . Furthermore, since bureaucratization is
p o sitiv ely associated with income and education, variables whose effect
is to lower f e r t i l i t y , i t is necessary th a t th e ir effects be con
sidered also. However, the addition of persons of child-bearing age
is expected to ra ise f e r t i l i t y ; in f a c t, an e a r lie r study by
Marshall (1968) leads to the prediction th a t th is w ill be the primary
e ffe c t of migration. This discussion leads to the following
hypotheses:
(6a) Migration is d ire c tly related to f e r t i l i t y .
(6b) I f education, income, and bureaucratization are
co n tro lled , the magnitude of the association between
f e r t i l i t y and migration will be increased.
I t is argued below th a t migration is an important facto r in
disorganization of the black family. Consequently, not only will
th is facto r a ffe c t the proportions of young persons in the c h ild
bearing ages, as well as income and education, i t is also expected to
lead to disorganization of the black family. This discussion suggests
the following hypothesis which is sp ecific to the black community:
46
(6c) The magnitude of the d ire c t association between
migration and f e r t i l i t y will decrease i f family
disorganization is controlled.
Social Mobility
Data reported in Chapter II indicate th a t social m obility is
inversely associated with f e r t i l i t y . However, i t was also suggested
th a t social m obility, bureaucratization, and migration are closely
related phenomena. That i s , migrants tend to be white c o lla r workers
in the process of improving th e ir social position. At the aggregate
level the question is whether or not v ertical mobility has effects
upon f e r t i l i t y independent of migration and bureaucratization. Blau
and Duncan (1964), who attempted to answer th is question for
individuals, concluded th a t the independent effects of vertical
m obility are small.
The usual d efin itio n of social mobility presents a problem for
formulating hypotheses. Social mobility usually refers to the upward
or downward movement of individuals. Since c itie s are not closed
systems, and survey data are not av ailab le, i t is not possible to
measure th is concept d ire c tly . However, social mobility is an
important variable and an attempt is made to evaluate it.from an
ecological point of view, a t le a st in fe re n tia lly .
I t is argued th a t the movement of individuals is only one
aspect of social m obility. Also important is the extent to which a
given occupational stru ctu re is experiencing rapid change in it s
prestige composition. For example, a c ity in which the proportions
47
employed in high statu s occupations increases during some specified
time interval is probably very d iffe re n t from a c ity in which th is
proportion decreases. Such rapid changes r e f le c t a dynamic economic
stru ctu re in which the opportunities available to individuals are
e ith e r expanding or contracting. This is true regardless of whether
the growth or contraction of occupations is due to movement of
individuals within or from outside the system. This in te rp re ta tio n
assumes th a t extensive changes in the proportions in high or low
statu s occupations will a l te r general perceptions of the occupational
stru c tu re . Consequently, changes in the proportions of individuals a t
various status levels are lik e ly to a l te r the context, or "atmosphere"
within which decisions about f e r t i l i t y are made.
The im plications of such a context fo r f e r t i l i t y are complex.
One such im plication was suggested by Mosca (referenced in Freedman,
1961-1962). He argued th a t when social m obility is pervasive even
non-mobile individuals are affected since they may have to lim it th e ir
family size merely to maintain a fixed position in the social order.
This argument suggests th a t rapid changes in the proportions in high
and/or low status occupations will be inversely associated with
f e r t i l i t y .
Since migration and occupational d iffe re n tia tio n are probably
closely associated with these s itu a tio n s , i t is necessary th a t they
be controlled. One argument, reviewed above, is th a t migrants are
in the process of improving th e ir s ta tu s . Migrants are also
concentrated in the child-bearing ages, and consequently tend to
48
increase the average level of f e r t i l i t y . This suggests th a t i f
migration is con tro lled , the association between social m obility will
be increased.
Sim ilarly, rapid change in .th e composition of an occupational
stru ctu re is lik e ly to lead to changes in the division of labor.
Although there are several dimensions to the concept "division of
labor," d iffe re n tia tio n is emphasized. The degree of d iffe re n tia tio n
re fe rs to the extent of differences among individuals in th e ir
sustenance a c tiv itie s . This depends upon the number of d iffe re n t
occupations and the d is trib u tio n of individuals among them. The
degree of d iffe re n tia tio n w ill therefore increase as the number of
occupations increases and as individuals become more evenly d istrib u ted
among them (cf. Labovitz, 1963; and Gibbs and Martin, 1962).
D ifferen tiatio n is expected to have consequences for systems.
Especially relevant are i t s effe c ts upon bureaucratization and
migration. A highly d iffe re n tia te d stru ctu re is lik e ly to require
bureaucratic stru ctu re s to coordinate the diverse functions implied
by i t . Sim ilarly, i t is expected th a t d iffe re n tia tio n is positively
associated with migration since an interchange of specialized
individuals among highly d iffe re n tia te d systems may be required to
sustain them.
Of course, changes in d iffe re n tia tio n are lik e ly to have other
consequences fo r urban stru c tu re s. However, these cannot be a n tic i
pated given the present s ta te of ecological knowledge, and i t is
therefore not possible to formulate other hypotheses. I t is necessary
49
th a t migration and d iffe re n tia tio n be controlled in order to evaluate
the independent effects of changes in the occupational stru ctu re .
This suggests the following hypothesis:
(7) The degree to which an occupational stru ctu re is
dynamic is inversely related to f e r t i l i t y i f
d iffe re n tia tio n and migration are controlled.
Family Structure
Although a p o te n tially important fa c to r affecting f e r t i l i t y ,
family stru ctu re has received re la tiv e ly l i t t l e a tte n tio n . An
exception is Petersen (1960), who argued th a t f e r t i l i t y rose sharply
during the in itia l phase of the demographic tra n sitio n in the
Netherlands due to the breakdown of tra d itio n a l controls over
f e r t i l i t y . A p art of th is breakdown was due to disruption of the
family stru ctu re . As a new family stru c tu re , and other in s titu tio n a l
controls emerged, f e r t i l i t y declined.
Numerous scholars have discussed the weakness of th is
in s titu tio n in the black community. For example, Farley (1965)
linked high black f e r t i l i t y to the family stru c tu re , as th is i n s ti tu
tion was affected by slavery and migration. He pointed out th a t
family disorganization began with the forced removal of blacks from
th e ir trib e s in A frica, coupled with forced association with members
of other trib e s . Furthermore, stab le fam ilies could not be
established in the United States since in no Southern s ta te could
slaves marry and couples could be separated with the sale of one
member.
50
Farley remarked th a t: "In the absence of both d is tin c tiv e
in s titu tio n s to regulate sexual and m arital practices and b irth control
techniques, high f e r t i l i t y would be predicted and the b irth rate
might approximate f if ty " (1965:397). Although marriage became legal
a f te r the Civil War, i t does not necessarily follow th a t a stab le
system was estab lish ed . Rather, the data suggest th a t a m atriarchial
system of "serial monogamy" was continued. Consequently, the re s u lt
was a continuation of the patterns of minimal control over f e r t i l i t y .
Also important was the rapid urbanization of the black
population, coupled with changes in the economic base and high
m obility. These facto rs may have complicated family disorganization.
That i s , whatever s ta b i lit y might have been introduced by leg aliz atio n
of marriage and ab o litio n of slavery 'was- probably more than o ffs e t
by these new forces.
The decrease in black f e r t i l i t y , with i t s consequent conver
gence with white f e r t i l i t y p rio r to the th i r tie s may be p a rtia lly
a ttrib u ta b le to the gradual creation of stab le in s titu tio n s and
tra d itio n s in ghettos. The abrupt r is e a f te r 1947 may be due in p art
to new disorganization a risin g from the rapid influx of blacks during
the 1940*s. This discussion of family stru ctu re leads to the
following hypothesis:
(8) Family disorganization is d ire c tly related to f e r t i l i t y .
Segregation
A fa c to r which is expected to be s p e c ific a lly rela ted to black
f e r t i l i t y is segregation. There are a t le a st two dimensions to th is
51
concept: behavioral and re s id e n tia l. Behavioral segregation refers
to the separation of blacks from whites in terms of occupations and
other group asso ciatio n s, both formal and informal. That i s , behav
io ral segregation refers to the degree to which blacks perform d if
fe re n t jobs and p a rtic ip a te in d iffe re n t groups. Residential
segregation refers to living p attern s—blacks living in separate parts
of the c ity . Of course, i t is probable th a t these two dimensions will
be highly co rrelated . (In f a c t, a key assumption is th a t such a
co rrelatio n e x is ts , as is seen in Chapter IV.)
Concerning segregation and f e r t i l i t y , i t is argued th at
iso la tio n of blacks from the remainder of the community will a ffe c t
th e ir f e r t i l i t y . This iso la tio n prevents assim ilation by blacks of
the re la tiv e ly low f e r t i l i t y norms of the to ta l community. Con
sequently, blacks w ill continue to adhere to the high f e r t i l i t y norms
brought from the rural south. This suggests the following
hypothesis:
(9) Segregation is d ire c tly related to black f e r t i l i t y .
CHAPTER IV
SOURCES OF D A T A A N D M ETH O D S OF ANALYSIS
Sources of Data
The data in the study are from two sources. The f i r s t source
is a data bank made available by Herman Turk, Director of the
Laboratory for Organizational Research a t the U niversity of Southern
C alifornia. Six variables rela tin g to the 130 United States c itie s
of more than 100,000 population in 1960 were taken from th is source.
The second data source fo r the 130 c itie s is the 1960 United States
Bureau of the Census reports (United States Bureau of the Census,
1963).
C ities of 100,000 were chosen because th is population size
seems to be the minimum a t which an urban place is meaningfully termed
a "c ity ." I t is f e l t th a t below th is siz e , urban places are less
lik e ly to possess the organizational a ttrib u te s usually thought of as
"urban." Although a rb itra ry , th is decision is consistent with
research which suggests th a t size is correlated with various urban
organizational p roperties. For example, Keyes (1958) showed th a t not
only is size correlated with organizational p ro p erties, there seem
to be plateaus in size a t which urban places possess many more
ch a ra c te ristic s than places below th a t siz e . One-hundred thousand
was found by Keyes to be a p artic u la rly important population siz e ;
52
53
c itie s in th is size class possess v irtu a lly a ll of the in stitu tio n a l
features considered in his study, such as r e ta il o u tlets and
specialty., shops.
Temporal Sequence
Since experimentation is impossible, any attempt to estab lish
relationships and causal chains must rely upon longitudinal and
cross-sectional observations. In the former, n aturally occurring
changes and th e ir apparent consequences are observed over time. In
the l a t t e r , observations are made a t roughly one point in time and i t
is assumed th a t some units have undergone a change which others have
not. That i s , an attempt is made to in fe r rela tio n s from the
observation of d iffe re n t magnitudes of v aria b les, and i t is presumed
th a t these differences are comparable to application of d iffe re n t
magnitudes of an independent variable.
There are s ta t is tic a l advantages and disadvantages to both
types of design. The principal advantage of longitudinal studies is
th a t some temporal rela tio n s can be em pirically established. However,
when regression and correlation coefficients are estimated from time
series data, internal autocorrelation becomes a potential problem in
both inference and in te rp re ta tio n . Autocorrelation occurs because
the observations in a time series are not independent of one-another;
under such circumstances, there is a tendency for erro r terms in
contiguous time periods to be co rrelated . In other words, i t is not
possible to assume th a t erro r terms are uncorrelated, a key assumption
in both regression and correlation analysis. The re s u lt is a certain
54
amount of " b u ilt-in ," and therefore spurious, co rrelatio n . Such
autocorrelations may be quite high (Kirk, 1960; Blalock, 1961, 1968).
With cross-sectional data dependence between observations and
e rro r terms is usually avoided and consequently more accurate
estim ates of correlation and regression co efficien ts are obtained. Of
course, i f the units are physically contiguous, a closely related
phenomenon can occur, as when patterns diffuse from one unit to
another (Blalock, 1968). Partly in consideration of these fa c to rs,
and prim arily because of a c c e sib ility of d ata, cross-sectional data
are used in th is study.
Ecological Correlations
The correlations used in th is study are based upon the
a ttrib u te s of groups. Consequently, no inference to the behavior of
individuals is intended. Such ecological correlations are interpreted
as indicators of more complex phenomena. For example, employment of
women is interpreted as refle ctin g the degree to which women are
integrated into society and trad itio n al tie s to the family are broken.
Operational Measures of Variables
Because i t is not possible to measure d ire c tly all of the
variables discussed in Chapter I I I , empirical indicators of certain
variables are used. Since an indicator is assumed to be correlated
with the variable of in te r e s t, i t consequently may account fo r only
part of the variance in the l a t t e r (Lazsarsfeld, 1960). Curtis and
Jackson (1962) advocate use of m ultiple indicators of concepts for
which no single unambiguous measure e x ists. An important advantage
55
cited by these authors is the greater control over spuriousness; th a t
i s , i f two indicators of the independent variable are rela ted in the
hypothesized way to a dependent v aria b le, there is less chance th a t
covariation re s u lts from some th ird v ariab le. If some of the
hypotheses are not supported, then ex post facto analysis is f a c i l i
tated as w ell; for example, i f one in d icato r is not rela ted as
hypothesized to the dependent v ariab le, and a second in d icato r i s ,
then i t is possible to compare the two in order to ascertain what i t
is about the concept which led to the fa ilu re in precision. A summary
l i s t of indicators and sources is presented in Table 1.
Income
Two measures of income are used, both taken from Census
publications: per cent of fam ilies with to ta l income of less than
$3,000 and per cent of fam ilies with to ta l income greater than or
equal to $10,000. A "family" is defined as "two or more persons
living in the same household who are related to each other by blood,
marriage, or adoption" (United States Bureau of the Census, 1963).
The income of fam ilies rath e r than individuals was chosen because th is
u n it is the one most lik e ly to be involved in decisions about children.
The per cent of fam ilies with to tal income less than $3,000,
or greater than or equal to $10,000, were chosen because these are
extremes which appear to represent "deprivation" and "affluence."
C ities with a large proportion of th e ir population concentrated in the
"affluent" category are lik e ly to be very d iffe re n t from c i tie s with
concentrations in the "deprived" category. Furthermore, the amount of
56
TABLE 1
INDEXES A N D SOURCES
Variable
Low income - to tal
1
High income - to tal
1
Low income - nonwhite^
Density
LFPR women - to ta l
1
LFPR women - to tal "bureau
c ra tic "
LFPR women - nonwhite
LFPR women - nonwhite
"bureaucratic"1
1
Low education - to ta l
1
High education - to tal
1
Description
Per cent of a ll fam ilies with 1959
income of less than $3,000
Per cent of a ll fam ilies with 1959
income greater than or equal to
$10,000
Per cent of nonwhite fam ilies with
1959 income of less than $3,000
Number of persons per square mile
Total number of employed females
divided by the to ta l number of
women 14+
Total number of females employed
in bureaucratic" occupations
(professional, te ch n ical, kindred;
managers, o f f ic i a ls , p ro p rieto rs;
c le r ic a l; and sales) divided by
to ta l number of women 14+
Total number of employed nonwhite
females divided by to ta l number of
nonwhite females 14+
Total number of nonwhite females
employed in "bureaucratic" occupa
tions (professional, te ch n ica l, and
kindred; managers, o f f ic i a ls , and
pro p rieto rs; c le r ic a l; and sales)
divided by the to tal number of
nonwhite females 14+
Per cent of all persons with less
than five years of education
Per cent of a ll persons 25+ with
greater than or equal to a college
education
57
TABLE .1—Continued
Variable
Low education - nonwhite1
High education - nonwhite1
Migration - to ta l'
Mobility - to ta l'
Migration - nonwhite1
Total v ertical m obility1
Occupational m obility1
Description
Per cent of nonwhite persons with
less than five years of education
Per cent of nonwhite persons 25+
with g reater than or equal to a
college education
Per cent of population five years
and older who resided in a d iffe re n t
county in 1955 as in 1960
Per cent of the population five
years and older who moved into
th e ir residence between 1958 and
1960
Per cent of nonwhite population five
years and older who resided in a
d iffe re n t county in 1955 as in 1960
Index of d issim ila rity for nine
major occupational categories;
1950 compared with 1960. Following
occupational categories used:
professional, te ch n ica l, and
kindred; managers, o f f ic i a ls , and
p ro p rieto rs; c le r ic a l; sa le s;
craftsmen, foremen, and kindred;
operatives and kindred; private
household workers; la b o re rs, exc.
farm and mine
Per cent of population in "high
statu s" occupations in 1960 minus
per cent of population in "high
statu s" occupations in 1950.
"High statu s" defined as: profes
sio n al, technical, kindred; and
managers, o f f ic ia ls , and
proprietors
58
TABLE ;1 —Continued
Variable
Occupational d iffe re n tia tio n -
t o t a l 1
Occupational d iffe re n tia tio n -
nonwhi te 1
Bureaucratization - t o t a l 1
Bureaucratization - nonwhite1
3
Segregation
Descri pti on
. Index of d is sim ila rity for the
to tal population employed in:
professional, tech n ica l, and kin
dred; managers, o f f ic i a ls , and
p ro p rieto rs; c le r ic a l; s a le s;
craftsm en, foremen, and kindred;
o p eratives, and kindred; private
household workers; la b o rers, except
farm and mine
Index of d iss im ila rity fo r the
nonwhite population employed in:
professional, te ch n ica l, and
kindred; managers, o f f ic i a ls , and
p ro p rieto rs; c le r ic a l ; sa le s;
craftsmen, foremen, and kindred;
o peratives, and kindred; private
household workers; la b o re rs,
except farm and mine
Total number of males and females
employed in "bureaucratic" occupa
tions divided by the to ta l number
of employed. "Bureaycratization"
is defined as employment in:
p rofessional, te ch n ica l, and
kindred; managers, o f f ic i a ls , and
p ro p rieto rs; c le ric a l and kindred;
and sales
Total number of nonwhite males and
females employed in "bureaucratic"
occupations divided by the to tal
number of employed nonwhites.
"Bureaucratization" is defined as
employment in: p rofessional,
tech n ica l, and kindred; managers,
o f f ic i a ls , and p ro p rieto rs; c l e r i
cal and kindred; and sales
Index fo r c i tie s computed on the
basis of block data
TABLE 1—Continued
5 9
Variable
Family disorganization - to t a l1
Family disorganization -
nonwhi tel
F e r tility r a tio 1
CEB25-341
Description
Divorced, separated, and widowed
females divided by to ta l number of
married females 14+
Divorced, separated, and widowed
nonwhites divided by to tal number
of nonwhite married females 14+
Children aged 0-4 divided by to tal
number of women aged 15-49
Children ever born per 1,000 ever
married women aged 25-34
United States Bureau of the Census, United States Census
of Population: 1960. Vol. 1, C haracteristics of the Population.
Washington, D.C.: United States Bureau of the Census, 1963.
- ?
United States Bureau of the Census. County and City Data
Book, 1962. Washington, D.C.: United States Bureau of the Census,
1962.
3
Taueber, Karl and Alma Taueber. Negroes in C itie s. Chicago:
Aldine, 1965.
60
income which can be diverted away from such necessities as food and
s h e lte r are also lik ely to be very d iffe re n t in these two types of
populations.
Cost
Density is the principal dimension of cost considered in th is
study, because under the condition of high density, l i t t l e space is
available and consequently i t becomes more expensive. This variable
is measured as the ra tio of number of square miles within the p o litic a l
boundaries of a c ity to i t s to ta l population. I t is argued th a t th is
ra tio d ire c tly re fle c ts the density of settlem ent.
Integration of W om en
Integration of women was defined in Chapter III as emancipation
of women from strong family t i e s ; i t was argued th at th is is best
reflected in the extent to which women p articip a te in the labor force.
Two indicators of labor force p articip atio n of women are used in the
study. The f i r s t , to tal labor force p articip a tio n of women, is
operationally defined by the following ra tio :
to ta l number of employed women 14+
to ta l number of women 14+
A second and more sp ecific operational d efin itio n involves
the following ra tio :
61
to ta l number of women 14+ employed in
sa le s , c l e r ic a l, managerial, o f f ic i a l, and
professional work
to ta l number of females 14+
The second ra tio permits more detailed an aly sis, since i t
re fe rs to women employed in "bureaucratic" kinds of jobs. I t has been
argued ( e .g ., Stycos, 1968) th a t employment of women in service jo b s,
especially "private household," has l i t t l e impact upon f e r t i l i t y
because l i t t l e role c o n flic t is involved. Children, for example, can
be supervised on the jo b , and women employed in non-bureaucratic jobs
are more lik e ly to possess a Gemeinschaft o rie n ta tio n , and consequently
may tend to have larg er fam ilies. I t is anticipated th a t th is
measure will be especially relevant fo r nonwhite population, because
such a large proportion have tra d itio n a lly been employed in non-
bureaucratic kinds of jo b s, especially service.
Education
Education is measured by the number of years of formal
schooling. A problem with th is measure, fo r comparative purposes, is
the substantial regional and racial variatio n in the qu ality of
schools. I t is doubtful th a t a diploma from a predominantly black
high school located in a Southern s ta te is comparable to one from a
Northeastern suburban high school. However, number of years of formal
schooling is the only measure of education available.
Because of the above problems in in terp retin g th is measure,
the proportions concentrated in the Census categories "less than five
62
years of schooling," and "greater than or equal to college graduation"
are used. The former category c learly represents a very poorly
educated group. I t is doubtful i f individuals in th is category possess
any more than the basic minimum of s k ills and a ttitu d e s customarily
associated with the concept of "education," regardless of the section
of the country or racial composition of the school from which i t was
obtained. S im ilarly, those individuals who have graduated from college
probably possess a re la tiv e ly large number of such s k ills and
a ttitu d e s --a g a in , regardless of region or racial balance of school.
Migration
The movement of individuals into and within the urban system
is measured in two ways. The primary measure is the per cent of a
c i ty 's population which reported th e ir residence to have been in a
d iffe re n t United States county in 1955 than in 1960. This closely
corresponds to the meaning customarily attached to the concept
"migrant"—th a t i s , a person who has changed both his occupational and
associational groups (cf. Hawley, 1950). However, th is measure
underestimates the to ta l migration ra te since i t l i s t s as non-migrant
persons who moved into the c ity from a d iffe re n t county between 1956
and 1960. On the other hand, i f i t is assumed th a t migration rates
in the years which intervened between 1955 and 1960 are correlated
with the ra te in 1955, then nothing would be gained by adding th is
information to the index.
Because of th is ambiguity, a measure of to tal volume of
movement also used—the per cent of the c i ty 's population which had
63
moved into th e ir present house between 1958 and 1960. A problem with
th is index is th a t not a ll such movers are "migrants" in the sense
defined above. That i s , many persons included in th is percentage are
intraurban movers who changed n eith er th e ir occupational nor other
associational groups. This measure is not available fo r the nonwhite
populations of the c itie s studied.
Social Mobility
I t was argued in Chapter III th a t a dimension of social
m obility relevant to urban stru ctu re s is the amount of changes in the
proportions employed in high and low sta tu s occupations. Such changes
a l te r the extent to which the occupational stru c tu re is perceived as
open or closed and th a t even the f e r t i l i t y of non-mobile individuals
is affected i f m obility is extensive.
This d e fin itio n requires a measure of the absolute amount of
change in the occupational stru c tu re a t two points in time. A
situ a tio n in which there are extensive changes in the proportions in
high statu s occupations, and an associated decrease in the proportions
in low statu s occupations, is illu s tr a te d in Table 2 by City "B." In
City "A," on the other hand, the proportionate d istrib u tio n remains
constant between t^ and tg . I t 1s argued th a t the occupational
stru ctu re of B is much more dynamic than th a t of A, and corresponds
to the d e fin itio n of a "mobility context" defined in Chapter I I I .
A simple measure, the Index of D issim ilarity (ID), is used
to evaluate these situ a tio n s (cf. Labovitz, 1963). The ID is defined
as: ( |X - Y |)/2 where X = per cent of population in a given
TABLE 2
EX A M PLES OF MOBILITY CON TEXTS
City A City B
Rank order of
occupations
by prestige
Proportionate d istrib u tio n
of population in occupa
tions a t:
t , - t 2 proportionate d istrib u tio n of
population in occupations a t:
*1 " *2
*1 *2 t l t2
1 .2 .2 .00 .2 .10 .10
2 .2 .2
o
o
•
.2 .10 .10
3 .2 .2 .00 .2 .05 .15
4 .2 .2 .00 .2 .25 •
o
CJI
5 .2 .2
o
o
•
.2 .50 .30
o>
65
occupation a t t-j, and Y = per cent of the population in th a t
occupation a t tg (cf. Labovitz, 1963).
The value of the ID can range from 0 to 1.00 and i t has a
straightforw ard in te rp re ta tio n . I t refers to the proportion of the
population a t which would have to change i t s occupation to make
i t s d istrib u tio n the same as a t t-|. Furthermore, the ID is a
measure of the to tal amount of change in an occupational stru ctu re
between two points in tim e, because a ll occupational categories are
included in the index. The occupational categories upon which the
ID are based is : professional, te ch n ical, and kindred workers;
managers, o f f ic ia ls , and p ro p rieto rs, except farm; c le ric a l and
kindred; sales workers; craftsm en, foremen, and kindred; private
household workers; service workers, except private household; and
lab o rers, except farm and mine (United States Bureau of the Census,
1963).
Bureaucratization
In Chapter I I I , bureaucratization was defined as the degree
to which a G esellschaft ra th e r than a Gemeinschaft orientation
pervades a system. S p ecifically , the concern is prim arily with the
q uality or type of interpersonal relatio n sh ip s in a system. I t is
argued th a t in bureaucratic se ttin g s relationships tend to be in s tru
mental rath er than expressive; members tend to tr e a t people as means
to ends. This orien tatio n is presumed to carry over into decisions
about family size.
The members of some occupations are more lik e ly to have a
66
Gemeinschaft orientation than members of other occupations. In
p a rtic u la r, Blau and Duncan (1964) suggest th a t th is is true of'w hite
c o lla r workers. The following occupations are considered to be in the
"white co llar" category: p rofessional, te ch n ical, and kindred;
managers, o f f ic i a ls , and p ro p rieto rs, except farm; c le ric a l and
kindred; and sales. The operational d efin itio n of bureaucratization
is :
number employed in : professional, technical
and kindred; managers, o f f ic i a ls , and
p ro p rieto rs, except farm; c le ric a l and
kindred; sales
to ta l number employed
Occupational D ifferentiation
Defined as the extent of differences among individuals in
th e ir sustenance a c t iv it ie s , i t was argued in Chapter III th a t th is
concept has a t le a st two dimensions: number of occupations, and
d istrib u tio n of individuals across these occupations. As the number
of occupations within a system increases, g reater differences among
individuals becomes possible. However, the number of occupations
per se is not an adequate measure of occupational d iffe re n tia tio n .
For example, consider a system with twenty occupations in which 99
per cent of the population is concentrated in a single occupation;
such a system clearly has le ss occupational d iffe re n tia tio n than one
in which there are only five occupations but individuals are evenly
d istrib u ted across them. In th is regard, Labovitz (1963:73) notes th a t
67
1 1 . . . a n appropriate measure should re fle c t both the number of
occupations and the d istrib u tio n s of individuals among them." This
leads to the second dimension of occupational d iffe re n tia tio n —
d istrib u tio n of individuals among occupations. Given a fixed number
of occupations, systems in which individuals are more evenly d i s t r i
buted will have the greater occupational d iffe re n tia tio n . A measure
which re fle c ts these dimensions can be defined with the following
formula (Labovitz, 1963): D = (1 - (£X2 / (£X)2)) where D equals
occupational d iffe re n tia tio n and X equals the number of individuals
in each occupation or occupational category.
"D" is used as a measure of occupational d iffe re n tia tio n in
each c ity . Its value is computed for each c ity based upon the
following census categories: professional, tech n ical, and kindred
workers; managers, o f f ic i a ls , and p ro p rieto rs, except farm; clerical
and kindred workers; sales workers' service workers, except private
household; and lab o rers, except farm and mine (United States Bureau
of the Census, 1963).
According to Labovitz, th is measure has an important
c h a ra c te ristic which should be kept in mind. The relationship between
"D" and differences among individuals is not lin e a r. As such
differences increase from zero, larg er and larger changes are required
to produce a comparable change in the value of "D." In other words,
the relatio n sh ip between "D" and differences among individuals can be
represented with a negatively accelerated curve.
68
Segregation
The two important dimensions of the concept of se g re g a tio n -
behavioral and re s id e n tia l—are assumed to be highly co rrelated ,
although data are available only for the l a tte r . This assumption is
based upon the following logic. I t is doubtful i f blacks could be
placed in a separate part of the c ity to liv e , and y et be able to
p a rtic ip a te fu lly in occupational and other groups. Sim ilarly, i f
blacks are dispersed throughout the c ity , i t is also lik ely th a t they
will be integrated into other groups as w ell. In any case, i t is
assumed th a t the degree of residential segregation re fle c ts behavioral
segregation.
The Tauebers (1965) have computed a measure of segregation for
a number of United States c itie s based upon 1960 Census block data.
They reasoned th a t i f there were no segregation, blacks would be
represented in the same proportions on each block as in the c ity as
a whole. For example, i f blacks comprise one-half of a c ity 's
population, and i f there were no segregation, then i t is expected th a t
every block would be one-half black. I f there were complete segrega
tio n , then blacks would not liv e on any block where whites resided.
The Tauebers used the Index of D issim ilarity (defined above),
which equals zero when there is no segregation and 100 when there is
complete segregation. Its in terp retatio n is analogous to the
in terp reta tio n fo r the m obility measure. That i s , i t refers to the
per cent of blacks who would have to change the block on which they
live in order to produce a pattern in which blacks were found bn each
69
block in the same proportion as the c ity as a whole. The c i tie s and
ID values for each c ity are reproduced in the Tauebers' study (1965).
Segregation indexes were not computed by the Tauebers for
fourteen of the la rg e st United States c i tie s because fewer than two
and one-half per cent of th e ir populations were black. These same
c itie s are eliminated in a ll comparisons of black f e r t i l i t y with to tal
f e r t i l i t y . Furthermore, correlations are based upon data for the
nonwhite rath e r than the black populations of these c itie s (with the
exception of the segregation indicator) since the necessary d is tin c
tions are not made by the Census. Since 92 per cent of the nonwhite
population is black, i t is not anticipated th a t th is will seriously
bias conclusions.
Family Disorganization
There are several d efin itio n s of th is concept, ranging from
"unhappiness" to break-up of the m arital u n it by divorce or desertion.
The former is exceedingly d if f i c u lt to measure, and is not considered
here. Instead, the proportion of the married population fourteen
years and older who reported themselves as divorced or separated is
taken as an operational d e fin itio n of family disorganization.
F e rti1i ty
F inally, two operational measures of f e r t i l i t y are used. The
principal measure is the f e r t i l i t y r a tio , defined as the ra tio of
children aged 0 through 4 to women aged 15 through 49. The advantage
of th is measure is th a t i t refers to the f e r t i l i t y of the e n tire
community and i t is re s tric te d to the five year period immediately
70
preceding the 1960 Census. The disadvantage of the f e r t i l i t y ra tio
is th a t i t is affected by variations in age and m arital composition.
C ities with the same age-specific b irth rates may have very d iffe re n t
f e r t i l i t y ra tio s i f e ith e r the proportions of women in the c h ild
bearing ages or the proportions married d iffe r.
A second measure of f e r t i l i t y is used which sp ec ific a lly
controls for age and m arital composition. This is the number of
children ever born per 1,000 ever married women aged 25-34. Since
th is measure refers only to women who are currently married, or have
been married, i t is unaffected by variations in m arital composition.
Furthermore, since i t indicates the f e r t i l i t y of th is group per 1,000,
i t is also unaffected by variations in age composition.
A problem with the children ever born measure is the time
period to which i t re fe rs . Although many of the women in th is age
category are in the process of forming th e ir fam ilies, or have ju s t
completed a major portion of th is process, some of the children
included were born as early as 1945. Since the c h a ra c te ristic s
included in th is study may have changed since th is d ate, there is some
question as to i t s appropriateness. For example, the per cent of a
given c i ty 's population which had graduated from college may be very
d iffe re n t in 1960 than i t was in 1945. A rela ted problem is inferring
"cause" to a variable measured in 1960 which presumably is affectin g
b irth s in 1945. Consequently, the f e r t i l i t y ra tio is emphasized
except in those hypotheses where age and m arital stru ctu re are
p o te n tia lly confounding facto rs.
_______________________________________________________________________________
71
Methods of Analysis
This study has two principal goals: (1) exploring the
relatio n sh ip s between f e r t i l i t y and certain dimensions of urban
stru c tu re ; and (2) evaluation of hypotheses derived from theoretical
premises. The f i r s t goal focuses upon the magnitude of rela tio n sh ip s,
and questions of cause are unimportant. For example, from th is point
of view i t does not m atter whether labor force p articip atio n of women
causes low f e r t i l i t y or vice v ersa. Such empirical generalizations
are important because they are an essential p rerequisite to "causal"
theory.
The second goal requires concern with more than the degree of
association. Id eally , causal hypotheses would be evaluated with
so-called "structural models"--a s e t of equations which correspond to
actual causal processes. Examples of such models include "path
analysis" (see, fo r example: Duncan, 1966) and Blalock's "causal
models" (1961). The assumptions required by such models cannot be
completely met with available data and theory. In p a rtic u la r, i t is
not possible to unequivocally estab lish all causal orderings.
Consequently, concern here is with evaluation of "p a rtia l" or
"incomplete" causal systems. The theory outlined in Chapter III does
permit hypotheses about intervening mechanisms and statements about
"causal" re la tio n s . The concept of "cause" is one of the most
ambiguous in sociology and requires a theoretical ratio n ale which
orders variables and specifies th e ir rela tio n s to one another (cf.
Theodorsen, 1967). This ratio n ale ought to suggest answers to the
72
following questions: does X lead to Y d ire c tly , or is i t s e ffe c t
through some other variable? Is there a d ire c t and an in d ire c t
effect? Furthermore, th is ratio n ale should indicate which controls
are important since i t is not possible or meaningful to control for
every variable.
A goal which is related to the two primary goals is evaluation
of the over-all predictive efficiency of any given s e t of variables
and th e ir re la tiv e importance. This is p a rtia lly a predictive and
p a rtia lly a causal an aly sis. To the extent th a t focus is upon se le c
tion of the s e t of indices which combine to best predict f e r t i l i t y ,
then the concern is p redictive. If an attempt is made to evaluate
the causal im plications of th is s e t of v aria b les, and th e ir re la tiv e
importance, then the concern is "th e o re tic a l."
This phase of analysis is largely inductive. The discussion
of hypotheses in Chapter III did not include statements about optimal
predictive combinations or re la tiv e importance of variab les. I t is
anticipated th a t th is approach will suggest hypotheses which can be
subjected to te s t in other stu d ies. The mathematical models u tiliz e d
in th is aspect of the study are variants of m ultiple co rrelatio n and
regression, and are described below.
Pearsonian Product Moment C orrelation
B ivariate re la tio n s in th is study will be measured with the
Pearsonian product moment co rrelatio n c o e ffic ie n t, or " r." There are
a number of in te rp re ta tio n s of r , two of which will be stressed
here. The f i r s t is amount of variance in a dependent variable
73
"explained" by an independent variable. However, as noted by
Labovitz and Hagedorn (1969), the magnitude of a given b iv ariate
correlation is not e n tire ly satisfa c to ry as a c rite rio n of cause given
the concept of m ultiple causation.. They suggest th a t two variables
may be causally related even when the magnitude of association is
very low. Consequently, since a major assumption of th is study is
th a t f e r t i l i t y is determined by several fa c to rs, emphasis is also
given to the "rate of change" in terp reta tio n of r ; th a t i s , in
standard score form, the value of r is interpreted as the amount
of change produced in a dependent variable by a u n it change in an
independent variable (McNemar, 1962).
P artial Correlation
P artial correlation is a d ire c t extension of the biv ariate
correlation c o e ffic ie n t described above, and closely related to
standardized regression c o e ffic ie n ts. This co efficien t measures the
variance in an independent variable "explained" by an independent
variable a fte r the variance a ttrib u ta b le to one or more other variables
has been taken out (Dubois, 1957). The p artial co rrelatio n coefficient
is used to evaluate hypotheses regarding spuriousness, intervening
v ariab les, and in d irect e ffe c ts .
Multiple Correlation and Regression
Hypotheses concerning predictive efficiency and the re la tiv e
importance of variables are evaluated with m ultiple correlation and
m ultiple regression c o e ffic ie n ts. Multiple regression is a technique
in which weights are assigned to independent variables to obtain
74
maximum predictive efficien cy ; these weights are regression
co efficien ts. W hen regression coefficients based upon raw scores are
"standardized" they can be used to compare variables in terms of th e ir
re la tiv e predictive importance. Standardization refers to conversion
of a ll variables to a common u n it of measurement with means equal to
zero and standard deviations equal to one. This is accomplished by
subtracting each score from the mean score of th a t variable and
dividing by the standard deviation of a ll scores.
Such standard co efficien ts are interpreted as indicating the
amount of change produced in a dependent variable by a unit change
in an independent variable with other variables in the equation
controlled. In other words, i t is a p artial slope. These standarized
regression co efficien ts are often known as betas. Betas can also be
interpreted as variance explained; a given beta squared refers to the
amount of variance in a dependent variable explained by an independent
variable with the effects of other variables in the equation
controlled.* Either in te rp re ta tio n is appropriate fo r ordering
independent variables in terms of th e ir re la tiv e effects upon a
dependent variable (McNemar, 1962).
Multiple correlation is closely related to m ultiple regression
using standardized regression c o efficien ts. I t indicates the amount
★
This in terp reta tio n is derived from the following equation,
which suggests an a lte rn a te technique for evaluating hypotheses regard
ing in d ire c t e ffe c ts : S y* s The term contain
ing the correlation co e ffic ie n t represents the in d ire c t e ffe c t through
tohidh each independent variable has on the dependent variable through
th e ir effects upon each other (cf. Hershi and Selvin, 1967).
75
o f variance in a dependent variable “explained" by a given combination
of independent variables. An a lte rn a te in te rp re ta tio n is th a t the
m ultiple correlation c o e ffic ie n t re fle c ts the accuracy with which a
given s e t of independent variables predicts a dependent variable
(McNemar, 1962).
CHAPTER V
ANALYSIS A N D DISCUSSION
In th is chapter the hypotheses presented in Chapter II are
evaluated. As was an ticip a ted , the magnitudes of the correlation
co efficien ts tend not to be large. Interpretations are based upon
one major c rite rio n : the consistency with which measures of the
independent and dependent variables are re la te d . Even i f the
correlations are re la tiv e ly sm all, i f they are co n sisten t, hypotheses
are considered to be supported.
This in te rp re ta tio n is supported by the in terco rrela tio n s
between the two measures of f e r t i l i t y . In the to tal population
children ever born correlates .76 with the f e r t i l i t y r a tio . This
indicates th a t f ifty -e ig h t per cent of the variance in these two
indicators is shared. The common variance is even less fo r the
nonwhite population, where the correlation between the two measures is
.46; therefore, only twenty-one per cent of the variance in the
f e r t i l i t y ra tio can be attrib u ted to the children ever born measure.
The fa c t th a t the two indicators of f e r t i l i t y do not completely
overlap suggests th a t each re fle c ts a somewhat d iffe re n t dimension of
f e r t i l i t y . Consequently, i f correlations based upon both measures are
co n siste n t, greater support is given to hypotheses.
76
77
In terp re tatio n of p a rtia l co rrelatio n co efficien ts is more
complex. Several hypotheses p redict th a t a given zero-order re la tio n
ship will be "more" or "less" positive (or negative) when controls
are introduced. Consistency is the principal c rite rio n applied to
such hypotheses. I f the p a rtia l co rrelatio n s based upon a lte rn a te
measures decrease or increase as predicted, then the hypothesis is
considered supported.
Income and F e r tility
Total Population
Hypothesis (la) predicts th a t the zero-order co rrelatio n
between income and f e r t i l i t y is negative. I t is seen in Table 3 th a t
the co rrelatio n s between d iffe re n t measures of income and f e r t i l i t y are
generally consistent with th is hypothesis. Only the correlation
between low income and f e r t i l i t y is in sig n ific a n tly small. Although
the magnitudes of the co rrelatio n c o efficien ts reported in Table 3
are not la rg e , th e ir consistency supports hypothesis (la ).
According to hypothesis ( lb ) , i f "relevant" variables are
co n tro lled , the association between income and f e r t i l i t y is p o sitiv e.
I t was argued th a t the inverse zero-order re la tio n between income and
f e r t i l i t y re s u lts from the positive association of income with
variables which are inversely rela ted to f e r t i l i t y . I f these
variables are co n tro lled , a tendency will emerge fo r populations to
increase th e ir production of children as income increases. Relevant
variables are education, bureaucratization, and employment of women.
The appropriate th ird order p a rtia ls between measures of income
TABLE 3
ZERO A N D PARTIAL CORRELATION COEFFICIENTS B ETW EEN
M EA SU RES OF IN CO M E A N D FERTILITY
Total Population Nonwhite Population
F e r tility
ratio^
CEB25-342 F e r tility CEB25-34
ra tio
Zero-order co rrelatio n s between -.14
> $10,000^ and measures of
T e r t ili ty
-.35 -.10 -.33
P artial correlations between -.12
> $10,000 and measures of .
T e r t ili ty with bureaucratization
LFPRW,6 and > College6 controlled
-.30 .13 -.22
Zer-order correlations between .02
< $3,000' and measures of f e r t i l i t y
.18 -.01 .42
P artial co rrelatio n s between .02
< $3,000 and measures of f e r t i l i t y
with bureaucratization, LFPRW ,
and < 5 y rs. ed. controlled
.14 -.08 .35
^Number of children aged 0-4 divided by number of women aged 15-49. See Chapter IV fo r a
discussion of th is measure. This abbreviation is used in a ll following ta b les.
2 03
Children ever born to women ever married aged 25-34. See Chapter IV fo r a discussion of
th is measure. This abbreviation is used in a ll following ta b les.
TABLE 3—Continued
3
Per cent of fam ilies with 1959 income g reater than or equal to $10,000. See Chapter IV for
a discussion of th is measure. This abbreviation is used in a ll following ta b les.
4
Number of persons employed in "bureaucratic" occupations divided by the to ta l number employed.
See Chapter IV fo r a discussion of th is measure. This abbreviation is used in a ll following ta b le s.
5
Labor force p articip a tio n ra te of women. See Chapter IV fo r a discussion of th is measure.
This abbreviation is used in a ll following ta b les.
^Per cent of the population with four or more years of college. See Chapter IV fo r a
discussion of th is variable.
^Per cent of fam ilies with 1959 incomes of less than $3,000. See Chapter IV fo r a discussion
of th is variable.
O
Per cent of the population aged 25 years or older with less than 5 years of education. See
Chapter IV fo r a discussion of th is variable.
V O
80
and f e r t i l i t y do not support th is hypothesis. For example, i t is seen
in Table 3 th a t the correlation between high income and the f e r t i l i t y
ra tio declines from -.14 to -.12 when bureaucratization, education,
and employment of women are controlled. In no case does the use of
controls markedly a l te r the zero-order co rrelatio n s indicating th at
income has independent inverse effe c ts upon f e r t i l i t y in the to tal
population. Therefore, hypothesis (lb) is not supported.
Nonwhite Population
Considering the data for nonwhites, the pattern of correlations
in Table 3 is also consistent with hypothesis ( la ). With the
exception of the co rrelatio n between low income and the f e r t i l i t y
r a tio , the correlations are of moderate magnitude and in the predicted
d ire c tio n . Therefore, hypothesis (la) is supported fo r the nonwhite
population.
Regarding hypothesis (lb ), when th ird -o rd er p a rtia ls between
measures of f e r t i l i t y and income are computed, controlling on
bureaucratization, education, and employment of women, the data are
in co n sisten t. The p a rtia l between high income and the f e r t i l i t y
ra tio is .13, and between low income and the f e r t i l i t y ra tio i t is
-.0 8 . Although these changes are not la rg e , they are in the predicted
d irec tio n . However, when children ever born is the dependent
v aria b le, the p a rtia ls are not s ig n ific a n tly d iffe re n t from the zero-
order correlation co efficien ts (Table 3).
These data are d if f i c u lt to in te rp re t. In view of the small
magnitudes of the p a rtia l co rrelatio n s between income and the f e r t i l i t y
81
r a tio , and the fa c t th a t the p a rtia l correlations between income and
children ever born remain negative, hypothesis (lb) appears not to be
supported fo r nonwhites. However, i t does appear th a t the independent
contribution of income is less in the nonwhite than in the to tal
population.
In summary, these data suggest th a t income and f e r t i l i t y are
inversely related in both the to ta l and in the nonwhite populations.
The inverse correlations p e rs is t when variables which might be
expected to lead to a "spurious" negative correlation in the to tal
population are controlled. The data are not so clear in the nonwhite
population; when relevant variables are controlled, the association
between income and the f e r t i l i t y ra tio reverses, and th a t between
children ever born and income is attenuated. This suggests th a t the
independent contribution of income is s lig h tly less in the nonwhite
than in the to tal population.
Cost and F e r tility
Total Population
The operational measure of cost co rrelates strongly with both
measures of f e r t i l i t y . That i s , density co rrelates -.42 with
children ever born, and -.44 with the f e r t i l i t y ra tio . This is
consistent with hypothesis (2) and suggests th a t one response of
populations to the costs imposed by lim ited space is lowered f e r t i l i t y .
Nonwhite Population
The data for the nonwhite population are also consistent with
hypothesis (2). However, the magnitude of the association between
82
the two measures of nonwhite f e r t i l i t y and density are much smaller
than in the to tal population. The correlation of density with the
f e r t i l i t y ra tio is -.2 1 , and with children ever born i t is -.18.
In sh o rt, hypothesis (2) is supported fo r both the to ta l and
the nonwhite population since the data suggest th a t cost facto rs are
important determinants of the f e r t i l i t y of urban populations. The
fa c t th at the correlations are much higher for the to tal population
than for the nonwhite population suggests th a t cost factors are less
important in the la t t e r group.
Integration of W omen and F e r tility
Total Population
Hypothesis (3) predicts th a t as women become more integrated
into the society, th e ir f e r t i l i t y will decline. Measuring integration
by employment of women, the hypothesis is supported for the to tal
population. The correlation between employment of women and children
ever born is -.3 5 , and between employment of women and the f e r t i l i t y
r a tio , i t is -.40.
Nonwhite Population
For nonwhites the data give v irtu a lly no support to the
hypothesis. The correlations between employment of women and children
ever born is -.0 1 , and between employment of women and the f e r t i l i t y
r a tio , i t is -.09. Consequently, among nonwhites the employment of
women is apparently unrelated to f e r t i l i t y .
However, this may be due to the occupations in which nonwhites
are concentrated, as suggested in Chapter I I I . Concentration of
83
nonwhite females in occupations such as service workers and unskilled
laborers may involve neither ro le c o n flic t nor imply th e ir integration
into the broader society.
Consequently, a more adequate te s t of hypothesis (3 ), the
correlation between the per cent of nonwhite females fourteen years
of age and older who are employed in "bureaucratic" occupations and
f e r t i l i t y was computed. The co rrelatio n between th is measure and the
f e r t i l i t y ra tio is -.1 7 , and with children ever born, i t is -.10.
These re su lts o ffe r support to the hypothesis. There is some tendency
for nonwhite f e r t i l i t y to decrease as women become more integrated
into the occupational stru c tu re , but the relatio n sh ip is not marked.
In general, comparison of the re la tio n between work and
f e r t i l i t y in the to ta l and the nonwhite population suggests important
differences. In the to ta l population, the co rrelatio n between the
bureaucratic p a rtic ip a tio n measures and f e r t i l i t y are somewhat lower
than the comparable co rrelatio n between employment of women in all
occupations and the two measures of f e r t i l i t y . In the nonwhite group
exactly the opposite is tru e . This suggests th a t employment means
something d iffe re n t in the nonwhite community than in the to tal
community. In the l a t t e r , any kind of increase in labor force
p articip a tio n rate s of women seems to imply integration of women into
the broader system. This does not seem to be the case in the nonwhite
community, where p a rtic ip a tio n in the labor force ger s£ has no e ffe c t.
The extent to which women are integrated into bureaucratic occupations
seems somewhat more important.
84
Education and F e r tility
Total Population
Hypothesis (4a) predicts an inverse relatio n sh ip between
education and f e r t i l i t y . The correlations reported in Table 4 support
th is prediction. Although the correlation between low education and
the f e r t i l i t y ra tio is n eg lig ib le, the other zero-order correlations
are consistent. Only the correlation between high education and
children ever born is moderate in magnitude. However, the general
consistency supports the hypothesis.
Hypothesis (4b) implies th a t the "true" re la tio n between
education and f e r t i l i t y is suppressed by the effects of migration.
This la t t e r variable is expected to be po sitiv ely correlated with
both f e r t i l i t y and education. The data presented in Table 4 are
consistent with th is hypothesis. Although none of the changes are
g reat, the p a rtia l co rrelations between measures of education and
f e r t i l i t y are uniformly larg er than the comparable zero-order
co rrelatio n s.
Hypothesis (4c) suggests two in d ire c t effects which education
may have upon f e r t i l i t y . I t is expected to be positively correlated
with bureaucratization and income, both of which are negatively
correlated with f e r t i l i t y . Assuming th a t bureaucratization and income
occur a fte r the educational process, th is hypothesis indicates a
possible in terp reta tio n of the inverse correlation between education
and f e r t i l i t y .
The p a rtia l correlations between education and f e r t i l i t y
TABLE 4
ZERO A N D PARTIAL
M EA SU RES
CORRELATION COEFFICIENTS B ETW EEN
OF EDUCA TIO N A N D FERTILITY
Total Population Nonwhite Population
F e rti1i ty
r a ti o
CEB25-34 F e r tility
ra tio
CEB25-34
Zero-order correlations between > college
and measures of f e r t i l i t y
-.15 -.30 -.31 -.16
P artial correlations between > college
and measures of f e r t i l i t y witTT the
following controls:
migration -.25 -.33 -.37 -.16
> $10,000 and bureaucratization -.11 -.2 0 -.2 2 -.04
Zero-order correlations between < 5 y rs.
ed. and measures of f e r t i l i t y
.02 .17 -.09 .15
P artial correlations between < 5 y rs. ed.
and measures of f e r t i l i t y with the
following controls:
migration .15 .25 -.09 .15
< $3,000 and bureaucratization .02 .02 -.13 -.01
00
C J 1
86
controlling fo r bureaucratization and income are shown in Table 4.
The zero-order correlation is reduced to approximately zero in the
case of low education and both measures of f e r t i l i t y . However, the
correlation between low education and f e r t i l i t y was almost zero
before the controls were introduced. The p a rtia l correlations between
high education and both measures of f e r t i l i t y are both reduced
somewhat. In terp re tatio n is therefore d if f i c u lt. The data suggest
th a t a portion of the e ffe c t of education upon f e r t i l i t y is through
bureaucratization and income. However, because control of these
variables does not elim inate the correlation between education and
f e r t i l i t y , i t appears th a t education also has a d ire c t e ffe c t.
I t was argued th a t education is also inversely related to
f e r t i l i t y when th is relatio n sh ip is examined sp ec ific a lly for women.
Since data on the education of nonwhite women are unavailable, th is
hypothesis is tested fo r the to tal population only. S pecifically ,
hypothesis (4d) predicts th a t the education of women is inversely
rela ted to f e r t i l i t y i f migration is controlled. Migration is
expected to be p o sitiv ely associated with both the education of women
and with f e r t i l i t y , and therefore tends to mask the re la tio n between
them.
As is seen in Table 5, the p a rtia l correlations between the
education of women and f e r t i l i t y with migration controlled tend to be
in the predicted d ire c tio n . Although the magnitudes of these
co rrelatio n s are low, th e ir consistency supports hypothesis (4d).
F inally, according to hypothesis (4e), a principal e ffe c t of
TA BLE 5
ZERO-ORDER A N D PARTIAL CORRELATION COEFFICIENTS B ETW EEN
M EA SU RES OF FEM A LE EDUCA TIO N A N D FERTILITY
Total Population
F e r tility
ra tio
CEB25-34
Zero-order co rrelatio n s between > 1 year
collegel and measures of f e r t i l i t y -.05 -.15
P artial co rrelatio n s between > 1 year college
and measures of f e r t i l i t y witlf the following
controls:
migration -.15 -.17
migration and LFPRW -.06 -.09
2
Zero-order co rrelatio n s between < 5 years ed.
and measures of f e r t i l i t y .05 .17
P artial correlations between < 5 years ed. and
measures of f e r t i l i t y with the following controls:
migration .08 .19
migration and LFPRW .11 .04
^Per cent of females twenty-five years of age and older with four or more year of college.
See Chapter IV fo r a discussion of th is measure.
C O
• V i
^Per cent of females twenty-five years of age and older with less than fiv e years tif
education. See Chapter IV fo r a discussion of th is measure..
88
female education upon f e r t i l i t y is through the integration of women.
If th is variable is controlled, hypothesis (4e) predicts th a t the
magnitude of the association between education of women and f e r t i l i t y
will decrease. The data are not e n tire ly consistent with th is
hypothesis. The p artial correlation between high education of women
and both measures of f e r t i l i t y decreases sharply when employment of
women and migration are controlled. The correlation between low
education of women and the f e r t i l i t y ra tio actually increases
slig h tly when employment of women and migration are controlled. How
ever, the correlation between low education of women and children ever
born decreases from .19 to .04, which is consistent with hypothesis
(4e). Consequently, hypothesis (4e) is generally supported.
To rec a p itu la te , the data for the to tal and the nonwhite
population tend to be co n sisten t. They indicate th a t education is
inversely associated with f e r t i l i t y . The data also suggest th a t
migration acts to suppress the true rela tio n between education and
f e r t i l i t y ; when th is variable is controlled, the association between
education and f e r t i l i t y is moderate in magnitude, suggesting th a t th is
is an important variable affecting f e r t i l i t y in urban areas.
Since the correlation between education and f e r t i l i t y in the
to ta l population is unaffected when bureaucratization and income are
controlled, i t appears th a t the effe c ts of education are not through
these variables. However, in the nonwhite population the p artial
correlations between education and f e r t i l i t y with bureaucratization
and income controlled are smaller than the zero-order c o rrelatio n ,
89
suggesting th a t bureaucratization and income are paths through which
education affects f e r t i l i t y .
F inally, the effects of female education were examined for
the to tal population. Although the data indicate th a t f e r t i l i t y is
inversely related to education of women, the small magnitudes of the
co rrelations indicate th a t th is is not an important variable. This
finding is inconsistent with previous d ata, both ecological and
individual, which emphasizes the importance of female education.
Whatever e ffe c t which th is variable has upon f e r t i l i t y in the urban
system appears to be a t the higher educational levels.
Bureaucratization and F e r tility
Total Population
Bureaucratization is expected to be inversely associated with
f e r t i l i t y . I t was argued in Chapter II th a t the growth of bureaucratic
structures leads to a pervasive ra tio n a lity which finds p artial
expression in the planning of family size. As is seen in Table 6,
bureaucratization and f e r t i l i t y tend to be inversely related i f no
other variables are controlled.
I t is also anticipated th a t bureaucratization is positively
associated with m igration. Since the d ire c t e ffe c t of bureaucratiza
tion is to lower f e r t i l i t y , while th a t of migration is to ra is e i t ,
i t 1s necessary to control the l a t t e r v ariab le. Hypothesis (5a)
predicts th a t the magnitude o f the inverse correlation between
bureaucratization and f e r t i l i t y w ill be increased i f migration is
controlled. I t is seen in Table 6 th a t th is hypothesis is given clear
TABLE 6
ZERO-ORDER A N D PARTIAL CORRELATION COEFFICIENTS B ETW EEN
BUREAUCRATIZATION A N D M EA SURES OF FERTILITY
Total Population
F e r tility
ra tio
CEB25-34
Nonwhite Population
F e r tility
ra tio
CEB25-34
Zero-order correlation between bureaucrati
zation and measures of f e r t i l i t y
P artial correlations between bureaucratiza
tion and measures of f e r t i l i t y with the
following variables controlled:
migration
> college, > _ $10,000
< 5 y rs. e d ., < $3,000
> co lleg e, > _ $10,000, migration
< 5 y rs. e d ., < $3,000, migration
-.06
-.30
.05
-.04
-.20
-.27
-.14
-.20
.11
-.08
.03
-.13
-.24
.30
.14
.28
.18
.32
-.24
-.24
.04
-.07
.03
-.08
91
support. When migration is co n tro lled , the co rrelatio n s between
measures of bureaucratization and f e r t i l i t y ris e sharply.
Hypothesis (5b) predicts th a t i f income and education are
co n tro lled , the co rrelatio n between bureaucratization and f e r t i l i t y
is lowered. I t was argued in Chapter III th a t bureaucratization leads
to higher income levels which, in tu rn , lead to lower f e r t i l i t y .
Furthermore, education is a link in the path through which bureaucrati
zation is expected to a ffe c t f e r t i l i t y . I f these variables are
c o n tro lled , i t is anticipated th a t the magnitude of the inverse
association between bureaucratization and f e r t i l i t y w ill decrease.
The pattern of p a rtia l correlations is consistent with th is
hypothesis (Table 6). For example, the p a rtia l co rrelatio n between
bureaucratization and children ever born with high education and high
income controlled is .11--a s h i f t of .25 in the predicted d irec tio n .
F in ally , hypothesis (5c) predicts an inverse correlation
between bureaucratization and f e r t i l i t y i f income, education, and
migration are controlled. The pattern of p a rtia ls reported in Table 6
tend to confirm th is hypothesis. For example, the co rrelatio n between
bureaucratization and the f e r t i l i t y ra tio with low education, low
income, and migration controlled is -.2 7 . These data are consistent
with the th e o retical ratio n ale th a t bureaucratization has d ire c t
effe c ts upon f e r t i l i t y .
Nonwhite Population
The same conclusions are Indicated fo r the nonwhite population.
The zero-order co rrelatio n s between nonwhite bureaucratization and the
Filmed as received
without page(s)
UNIVERSITY MICROFILMS.
93
f e r t i l i t y ra tio and children ever born are both -.2 4 . When migration
is controlled, the p a rtia l co rrelatio n between bureaucratization and
children ever born is unchanged, while th a t between bureaucratization
and the f e r t i l i t y ra tio increases s lig h tly to -.3 0 . This is
consistent with hypothesis (5a). However, since the change is not
g re a t, the im plication is th a t migration is not an important suppressor
variable in the nonwhite population.
The pattern of p a rtia l correlations between the two measures
of f e r t i l i t y and bureaucratization with education and income controlled
is generally consistent w ith, and supports hypothesis (5b). Only the
p a rtia l between bureaucratization and the f e r t i l i t y ra tio with low
education and low income controlled f a ils to decrease; in f a c t, i t
increases s lig h tly .
Hypothesis (5c) is also supported by the data. The p a rtia l
correlations.between bureaucratization and f e r t i l i t y with income,
education, and migration controlled range from -.32 to .03. As
with the to ta l population, only the l a t t e r is inconsistent.
To summarize, data analyzed in th is section suggest th a t
bureaucratization affe c ts f e r t i l i t y in complex ways. I t is p art of
the path through which education affects f e r t i l i t y , and i t also
appears to have in d ire c t effe c ts through i t s association with Income.
The positive association between migration and both bureaucratization
and f e r t i l i t y tends to suppress the true association between these
l a t t e r variab les. F inally, the fa c t th a t the association between
vureaucratization and f e r t i l i t y remains moderate in magnitude even i f
94
education, income, and migration are controlled suggests th a t
bureaucratization makes an independent contribution to f e r t i l i t y .
There appear to be few differences in the pattern of p artial
and zero-order correlations between the to ta l and nonwhite populations.
However, one difference is the e ffe c t of m igration. This variable
appears to a ffe c t the rela tio n sh ip between bureaucratization and
f e r t i l i t y rath er strongly in the to tal population, but has l i t t l e
impact in the nonwhite population.
Migration and F e r tility
Total Population
Hypothesis (6a) predicts th a t migration is positively related
to f e r t i l i t y . The co rrelatio n s between migration and m obility and
the f e r t i l i t y ra tio are moderate in size (Table 7 ). However, these
measures are v irtu a lly unrelated to children ever born. This gives
strong support to the in te rp re ta tio n presented in Chapter I I . A
principal e ffe c t of migration is expected to be through the addition
of young persons of child-bearing ages to the population. The
children ever born measure is standardized for age composition, while
the f e r t i l i t y ra tio is not. The re la tiv e ly high correlation of the
f e r t i l i t y ra tio with migration supports th is in te rp re ta tio n . Given
the fa c t th a t children ever born is standardized fo r age, i t would
not be expected to be closely related to migration i f the principal
e ffe c t of the l a t t e r were through i t s Impact upon the age stru c tu re .
Migration 1s also expected to a ffe c t f e r t i l i t y through its
impact upon education, income, and bureaucratization. Since these
TABLE 7
ZERO-ORDER A N D PARTIAL CORRELATIONS B ETW EEN
M EA SU RES OF M IGRATION A N D FERTILITY
Total Population Nonwhite Population
F e r tility
r a ti o
CEB25-34 F e r tility
ra tio
CEB25-34
Zero-order correlations between migration
and measures of f e r t i l i t y .32 .05 .28 .02
P artial co rrelatio n s between migration and
measures of f e r t i l i t y with the following
controls:
< 5 y rs. e d ., < $3,000 and
bureaucrati zati on .45 .12 .31 .09
> college, > $10,000, and
bureaucratization .42 .11 .35 .06
Zero-order co rrelatio n s between mobility* and
measures of f e r t i l i t y .38 .09 .27 .13
P artial correlations between m obility and
measures of f e r t i l i t y with the following
controls:
< 5 y rs. e d ., < $3,000, and
bureaucrati zati on .44 .12 .28 .16
^ c o lle g e , > $10,000, and
bureaucrali zati on .47 .14 .30 .13
V e r cent of the population which reported liv in g in a d iffe re n t dwelling u n it in 1958 01
as in 1960. See Chapter IV fo r a discussion of th is measure.
96
variables are expected to lead to lower f e r t i l i t y , they tend to mask
the "true" relatio n sh ip between migration and f e r t i l i t y . Therefore,
hypothesis (6b) predicts th a t i f these variables are controlled, the
correlation between migration and f e r t i l i t y w ill be increased. The
p artial correlations in Table 7 are uniformly consistent with th is
prediction. The association between migration and the f e r t i l i t y
ra tio increases .10 and .13, depending upon the p a rtic u la r
combination of controls. Smaller, but nonetheless consistent increases
are obtained for the mobility measure.
Nonwhite Population
As with the to tal population, the data are consistent with
hypothesis (6a). The zero-order co rrelatio n s between the two
measures of horizontal movement and the f e r t i l i t y ra tio are not quite
as high as for the to tal population, but are s t i l l moderate in
magnitude. Furthermore, the zero-order co rrelatio n s between migration
and m obility and children ever born are also in s ig n ific a n t.
Regarding hypothesis (6b), which predicts th a t the p artial
correlation between migration and f e r t i l i t y w ill increase i f
bureaucratization and education are co n tro lled , the data are not so
clear. The th ird -o rd er p a rtia l between the two measures of migration
and the f e r t i l i t y ra tio increase only slig h tly when bureaucratization,
education, and income are controlled. The increases fo r the m obility
measure are especially small. Therefore, hypothesis (6b) is not
adequately supported.
F in ally , hypothesis (6c) predicts th a t the association between
97
migration and f e r t i l i t y w ill decrease i f family disorganization is
controlled in the nonwhite population. I t was anticipated th a t
migration leads to g reater family disorganization. This in turn leads
to less control over sexual behavior, and a higher f e r t i l i t y ra te . If
th is path is removed, i t is anticipated th a t a portion of m igration's
e ffe c t w ill be taken out. However, the p a rtia l co rrelatio n s between
migration and the two measures of f e r t i l i t y with family disorganiza
tion controlled are almost the same as the zero-order co rrelatio n s.
The p artial co rrelatio n between migration and children ever born is
.03, and between the f e r t i l i t y ra tio and migration i t is .26. The
im plications of th is finding for nonwhite populations w ill be
examined in the section on family disorganization.
In summary, these data suggest th a t migration is an important
facto r affectin g the f e r t i l i t y of urban populations. For the to ta l
population i t appears th a t m igration's association with bureaucratiza
tio n , education, and income a ct to mask i t s importance. This does not
appear to be the case in the nonwhite population, where the principal
e ffe c t of migration appears to be d ire c tly upon the age stru c tu re .
This is explained by the re la tiv e ly small re la tio n between nonwhite
migration and the control variables compared with the to tal
population.
Social Mobility and F e r tility
Total Population
I t was anticipated th a t the amount of change in occupational
stru ctu res (social m obility) is inversely related to f e r t i l i t y .
98
According to the co rrelatio n s presented in Table 8, there is a very
s lig h t inverse re la tio n between the two measures of f e r t i l i t y and th is
variable. However, since i t was expected th a t changes in the
occupational stru ctu re would be correlated with migration and d iffe re n
ti a ti o n , these variables are controlled to evaluate the independent
contribution of a ll change. I t is seen in Table 8 th a t changes 1n the
occupational stru ctu re are e ss e n tia lly uncorrelated with e ith e r
measure of f e r t i l i t y . Therefore, hypothesis (7) is unsupported for
the to ta l population.
Nonwhite Population
The pattern of zero-order and p a rtia l correlations are
e s se n tia lly the same for the nonwhite as fo r the to ta l population. The
zero-order co rrelations between the two measures of f e r t i l i t y and
changes are in the predicted d irectio n but of very small magnitude.
The p a rtia ls with d iffe re n tia tio n and migration controlled are also
very small. Therefore, hypothesis (7) is not supported for the
nonwhite population.
To re c a p itu la te , changes in occupational stru ctu re was intended
as an in d ire c t measure of social m obility. I t was assumed th a t when
such change is extensive, the degree to which the occupational
stru ctu re is perceived as "open" w ill be a lte re d . When the stru ctu re
is perceived as open, i t was expected th a t individuals will tend to
reduce th e ir f e r t i l i t y so as to maintain a competitive position.
However, the v a lid ity of th is measure is questioned. I t
requires a complex, and p o te n tially dubious, inference chain. F ir s t,
TABLE 8
ZERO-ORDER A N D PARTIAL CORRELATIONS BETW EEN SOCIAL
MOBILITY A N D M EA SU RES OF FERTILITY
Total Population Nonwhite Population
F e r tility CEB25-34 F e r tility CEB25-34
ra tio ra tio
Zero-order correlations between social
mobility*and measures of f e r t i l i t y -.10 -.10 -.13 -.05
P artial co rrelatio n s between social
ipobility and measures of f e r t i l i t y with
migration and occupational d iffe re n tia tio n 2
controlled -.07 .06 -.11 -.03
^Measure of occupational change between 1950 and 1960. See Chapter IV fo r a discussion of
th is measure. This abbreviation is used in a ll following ta b les.
2
Index of D issim ilarity based upon the d istrib u tio n of populations among occupations in 1960.
See Chapter IV fo r a discussion of th is measure.
V O
vo
100
i t requires the assumption th a t rapid changes in the proportions
employed in various occupations are perceived by the to ta l population.
I t fu rth e r requires th a t such perceptions w ill lead to a general
d efin itio n of "openness." And f in a lly , i t requires th a t a d efin itio n
of the stru ctu re as "open" or "closed" leads to reduction of f e r t i l i t y .
Such a complex chain is open to question a t any point.
Yet, th is critic ism is not sp ecific to the measure used here.
Any statement th a t upward and downward m obility w ill generate an
"atmosphere" which w ill a ffe c t f e r t i l i t y re s ts upon a sim ilar logic.
Consequently, since the pattern of reasoning employed here is close
to th a t d ire c tly re la tin g m obility to f e r t i l i t y , doubt is cast upon
the theoretical premise th a t "in stitu tio n a liz a tio n of social mobility"
affe c ts f e r t i l i t y . The t e s t used here ought to be a t le a s t
tan g en tially relevant.
Family Disorganization and F e r tility
Total Population
Hypothesis (8) predicts a d ire c t rela tio n between family
disorganization and f e r t i l i t y . The hypothesis is not supported by the
data. In f a c t, as is seen in Table 9, family disorganization is
inversely associated with f e r t i l i t y . The correlation between th is
variable and the f e r t i l i t y ra tio is -.3 4 , and with children ever
born, i t is -.13.
In an attempt to in te rp re t the unexpected inverse association
between family disorganization and f e r t i l i t y , the labor force
p articip a tio n ra te of women was controlled. The ratio n ale fo r th is
TA BLE 9
ZERO-0RDER A N D PARTIAL CORRELATIONS B ETW EEN FAM ILY
DISORGANIZATION A N D M EA SU RES OF FERTILITY
Total Population Nonwhite Population
F e r tility CEB25-34 F e r tility CEB25-34
ra tio ra tio
Zero-order correlations between family
disorganization^ and measures of f e r t i l i t y -.34 -.13 -.08 .08
P artial co rrelatio n s between family
disorganization and measures of f e r t i l i t y
with the following controls:
LFPRW -.21 .01 .00 .26
Female bureaucratization^ -.31 -.09 .14 .33
Migration -.32 -.13 -.05 .08
^Number of divorced and separated females divided by the number of married females. See
Chapter IV fo r a discussion of th is variable.
2
Number of females in "bureaucratic" occupations divided by the to ta l number of females
aged fourteen years and older. See Chapter IV fo r a discussion of th is v ariab le. This
abbreviation is used in a ll tables in which th is variable appears.
o
102
control 1s th a t a high level of family disorganization may lead to
higher labor force p articip a tio n rates of women as they are forced
onto the labor market. This in turn is expected to lower f e r t i l i t y .
The re su lts are consistent with the hypothesis. The co rrela
tion between family disorganization and the f e r t i l i t y ra tio with the
labor force p articip a tio n ra te of women controlled is -.2 1 . The
comparable p a rtia l between family disorganization and children ever
born is .01. This suggests th a t an important part of family
diso rg an izatio n 's e ffects upon f e r t i l i t y are through i t s e ffects upon
the employment of women. The independent effects appear to be small.
Nonwhite Population
Consistent with the re su lts for the to tal population, the
zero-order correlations between family disorganization and f e r t i l i t y
do not support the hypothesis. As seen in Table 9, th is variable
co rrelates -.08 with the f e r t i l i t y ra tio and .08 with children
ever born. The magnitudes of these co rrelations are small and
inconsistent with each other.
As indicated previously, i t was expected th a t th is variable
would be po sitiv ely correlated with family disorganization in the
nonwhite population. Heavy migration has supposedly been responsible
fo r the general disorganization of ghetto l i f e , and especially the
ghetto family. In f a c t, there 1s v irtu a lly no re la tio n between
migration and family disorganization. The correlation of -.12
r e f le c ts , i f anything, a s lig h t Inverse association. Because of th is
small correlation the relatio n sh ip between family disorganization and
103
f e r t i l i t y is v irtu a lly unaffected when migration is controlled.
The apparent lack of association between migration and family
disorganization is inconsistent with a g reat deal of speculation and
theory regarding the black family.. I t is apparent th a t whatever the
pattern has been in the past, these data suggest th a t the stru ctu re
of the black family is not adversely affected by migration today.
F inally, as with the to tal population, an attempt was made to
in te rp re t the low relationship between family disorganization and
f e r t i l i t y by controlling employment of women. As seen in Table 9,
the p artial correlation between the f e r t i l i t y ra tio and family
disorganization with employment of women controlled is .00, and the
co rrelatio n between children ever born with the same control is .26.
These co rrelations are d if f i c u lt to in te rp re t.
As noted e a r l ie r , the low correlations between th is variable
and f e r t i l i t y may be due to the employment of black women in
"tra d itio n a l" occupations. Consequently, pressures toward lower
f e r t i l i t y are not created. To te s t th is e ffe c t, employment of nonwhite
women in "bureaucratic" occupations was controlled. The p a rtia ls are
shown in Table 9. The relationship between the two measures of
f e r t i l i t y and family disorganization is po sitiv e. In f a c t, the general
pattern of correlations with employment of women in a ll occupations
or female bureaucratization controlled tends to support the conclusion
th a t the apparent lack of relationship is due to the "masking" effects
of these variables. Therefore, i t is argued th at the e ffe c t 0 i amily
disorganization upon f e r t i l i t y in the black community is to increase
f e r t i l i t y .
104
In summary, the rela tio n sh ip between family disorganization
and f e r t i l i t y is very d iffe re n t in the to ta l population than in the
nonwhite population. In the former fam ily,disorganization appears
unrelated to f e r t i l i t y i f i t s e ffects upon employment of women are
con tro lled . That i s , family disorganization leads to an increase in
employment of women which, in tu rn , leads to a lower level of
f e r t i l i t y . This suggests th a t the strength or weakness of the family
stru ctu re does not a ffe c t f e r t i l i t y in the to ta l population.
Considering the nonwhite population, a d iffe re n t pattern is
found. When female bureaucratization or employment of women are
uncontrolled, the rela tio n sh ip is weak and in c o n sisten t. When these
variables are held constant the pattern of co rrelatio n s suggests th a t
the independent contribution of family disorganization to f e r t i l i t y
is d ire c t and moderate in magnitude.
Segregation and F e r tility
The fin al hypothesis, re la tin g segregation to f e r t i l i t y , is
sp ecific to the nonwhite population. I t was argued th a t the iso la tio n
of blacks may prevent th e ir assim ilation of the re la tiv e ly low
f e r t i l i t y norms c h a ra c te ristic of the to ta l community. Hypothesis (9)
predicts th a t nonwhite f e r t i l i t y is d ire c tly rela ted to segregation.
The zero-order co rrelatio n s presented in Table 10 are co n sisten t with
th is hypothesis. The co rrelatio n between segregation and the f e r t i l i t y
ra tio is .10, while th a t between segregation and children ever born
is .37.
However, there remains the p o s s ib ility th a t these co rrelatio n s
TA BLE 10
ZERO-ORDER A N D PARTIAL CORRELATIONS B ETW EEN SEGREGATION
A N D M EA SU RES OF FERTILITY
Nonwhite Population
F e r tility
ra tio
CEB25-34
Zero-ordir correlation between segregation^
and measures of f e r t i l i t y .10 .37
P artial correlations between segregation and
measures of f e r t i l i t y with the following controls:
< $3,000, < 5 y rs. e d ., and
bureaucratization .00 .15
> $10,000, > colleg e, and
bureaucraTi zati on .04 .22
^Index of resid en tial segregation based upon block data. See Chapter IV fo r a discussion
of th is variable.
o
tn
106
are e ith e r spurious or due to the effects of segregation upon variables
other than assim ilation of low f e r t i l i t y norms. Although controls
were not specified in Chapter I I , income, education, and bureaucrati
zation are important. When these are co n tro lled , the relationship
between segregation and the f e r t i l i t y ra tio becomes zero (Table 10).
However, the correlation between children ever born and segregation
with these variables controlled remains slig h tly p o sitiv e.
These p a rtia ls are d if f i c u lt to in te rp re t. I t may be th a t
segregation has a small independent e ffe c t upon f e r t i l i t y , as indicated
by the p a rtia l co rrelatio n s of .15 and .22 between th is measure
and children ever borrr. However, most of the relatio n sh ip between
segregation and f e r t i l i t y is accounted for by i t s association with
education, income, and bureaucratization. This suggests th a t the
in terp reta tio n presented in Chapter II is inappropriate even though
the zero-order co rrelatio n s are in the predicted directio n and of
moderate magnitude.
The proper in te rp re ta tio n depends upon the causal ordering
of the variab les. I f education, income, and bureaucratization are con
sidered p rio r in time to segregation, then the co rrelatio n between
th is l a s t variable and f e r t i l i t y is spurious. However, the issue 1s
more complex. I t may be th a t segregation is actually a cause of these
variables. In th is case, i t would be appropriate to argue th a t
segregation 1s an important variable which acts upon f e r t i l i t y through
such facto rs as bureaucratization, education, and income.
I t is not possible to fu lly untangle the causal order among
variables such as segregation and income and bureaucratization. I t is
107
probable th a t there is a feedback e ffe c t. That i s , blacks with low
income and statu s are re s id e n tia lly iso la ted . Once iso la te d , th is
facto r i t s e l f prevents acquisition of higher sta tu s jobs and incomes
which would f a c i l i t a t e leaving the. ghetto.
Multiple Regression and Correlation Analysis
In th is section the various stru ctu ra l determinants of
f e r t i l i t y are examined simultaneously. M ultiple regression and
co rrelatio n co efficien ts are used to evaluate the re la tiv e importance
of d iffe re n t variables as well as th e ir overall predictive efficien cy .
The standardized regression co efficien ts of a selected se t of variables
from the to tal and nonwhite populations are also examined to provide
some indication of th e ir re la tiv e importance.
M ultiple regression and co rrelatio n analysis is intended to do
very d iffe re n t things than the b iv a riate analyses of hypotheses
presented above. The former is intended to evaluate whether or not a
given variable makes an important independent contribution to f e r t i l
i t y , and to provide a ranking of variables in these terms. Bivariate
analyses are intended to evaluate hypotheses about the process through
which independent variables a ffe c t f e r t i l i t y in conjunction with
certain th e o re tic a lly relevant controls. Furthermore, analysis 1n
th is section is prim arily exploratory and inductive, while th a t in the
preceding sections was prim arily deductive.
The beta co efficien ts presented in th is section are generally
consistent with the previous analysis of b iv ariate hypotheses.
Variables which emerge as important predictors of f e r t i l i t y in m ultiple
108
regression equations are also lik e ly to have high zero-order co rrela
tions with f e r t i l i t y . Multiple regression merely provides a convenient
way of ranking variables.
F in ally , analysis is largely confined to the f e r t i l i t y r a tio ,
for two reasons. F ir s t, analysis of b iv a riate hypotheses and controls
have led to sim ilar conclusions whether the f e r t i l i t y ra tio or
children ever born was used as the dependent v ariab le; furthermore,
there was no consistent tendency fo r one dependent variable to yield
higher correlation c o efficien ts than the other. Secondly, the
f e r t i l i t y ra tio is the most conceptually c lear dependent v ariab le,
and is therefore most appropriate to an inductive analysis.
Total Population
When all the variables used in the study are simultaneously
correlated with the two measures of f e r t i l i t y , moderately high
m ultiple correlation co efficien ts re s u lt.* When children ever born
is the dependent v aria b le, the m ultiple correlation co efficien ts are
.59 and .64. When the f e r t i l i t y ra tio is the dependent variab le,
the m ultiple correlation co efficien ts are .63 and .65. This
suggests th a t a substantial portion of the variance in f e r t i l i t y is
"explained" by the independent variables used in the study.
Standardized regression co efficien ts are presented in Table 11.
*Two m ultiple correlation co efficien ts are obtained for each
measure of f e r t i l i t y because s lig h tly d iffe re n t sets of independent
variables are used. The f i r s t se t includes low education and low
income, and a ll other variables used in the study. The second s e t
includes high education and high income, and a ll other variables used
in the s t u d y . ______________________________________________________
109
TA B LE 11
STANDARDIZED REGRESSION COEFFICIENTS IN REGRESSION
EQUATIONS W ITH THE FERTILITY RATIO AS TH E
DEPEND EN T VARIABLE—TOTAL POPULATION
Equation Number One Equation Number Two
Density -.30 Density -.36
Migration .30 Migration .35
LFPR W *.21 Bureaucratization -.23
Bureaucratization -.19 LFPRW -.22
Family Disorganization -.18 Family Disorganization -.21
College -.14 < $3,000 -.18
> $10,000 -.09 < 5 y rs . ed. .20
Social Mobility -.06 Social Mobility ,07
no
Since d iffe re n tia tio n is so closely correlated with social m obility,
the former variable has been omitted. Furthermore, even though
measures of education and income are closely co rrelated , i t was decided
to include both measures because they are conceptually d is tin c t.
However, female education is excluded because of i t s conceptual
sim ila rity to to tal education measures. The same pattern is followed
in m ultiple regression analysis of nonwhite f e r t i l i t y .
Considering the two sets of equations presented in Table 11,
i t is seen th a t the rankings of beta co efficien ts are very sim ilar.
The only difference between the two is the position of female
employment and bureaucratization. However, the differences are
s lig h t.
Several important im plications can be derived from these
equations. F ir s t, i t is apparent th a t three more or less d is tin c t
"levels" of variables are found in terms of independent contributions
to f e r t i l i t y . Density and migration emerge as the variables making
the g reatest contribution to to ta l f e r t i l i t y . I t is seen th a t the
magnitudes of the co efficien ts are much larg er than those of th e ir
nearest com petitors.
Also important are female employment, bureaucratization,
education, and family disorganization. The beta co efficien ts for
these variables are approximately equal in magnitude, suggesting th a t
th e ir independent contributions to f e r t i l i t y are comparable. The
beta c o e ffic ie n t fo r social m obility is very small in both equations,
suggesting th a t i t is an unimportant variab le. The beta co efficien ts
Ill
for income are inconsistent; high income appears to be a re la tiv e ly
unimportant v ariab le, while low income emerges as moderately
important.
Nonwhite Population
Two sets of beta co efficien ts are presented for the nonwhite
population. The f i r s t s e t is based upon the same se t of independent
variables as were used to evaluate to ta l f e r t i l i t y . This permits
d ire c t comparison of the two groups. The second s e t includes these
variables plus segregation.
The magnitudes of the m ultiple co rrelatio n coefficients are
not quite as large in the nonwhite population as in the to ta l
population. W hen children ever born is the dependent v ariab le, the
m ultiple correlation co efficien ts are .49 and .46. When the
f e r t i l i t y ra tio is the dependent v ariab le, the m ultiple correlation
co efficien ts are .57 and .62. The magnitude of these co rrelations
suggest th a t a substantial amount of the variance in f e r t i l i t y is
"explained" by the independent variables used in the study.
Considering the variables in Table 12, the ordering of
variables between the two sets of equations tends to be co n sisten t.
Employment of women and family disorganization emerge as the strongest
contributors in both equations. Furthermore, i t is seen th a t high
Income and high education are also important. Less important, but with
beta co efficien ts of moderate magnitude, are m igration, bureaucratiza
tio n , and density. Relatively small independent contributions are
made by low income and low education. F inally, as with the to tal
112
TABLE 12
STANDARDIZED REGRESSION COEFFICIENTS IN REGRESSION
EQUATIONS W ITH THE FERTILITY RATIO AS THE
DEPENDENT VARIABLE—N O N W H ITE POPULATION
Equation Number One Equation Number Two
LFPRW -.43 LFPRW -.43
> College -.40 Family Disorgani
zation .38
Family Disorgani
zation .38 Bureaucratization -.30
> $10,000 .36 Migration .29
Migration .32 Density -.21
Density -.28 < $3,000 -.18
Bureaucratization -.25 < 5 y rs . ed. -.11
Social Mobility -.1 0 Social Mobility -.1 0
Equation Number Three Equation Number Four
> College -.44 LFPRW -.47
LFPRW -.44 Family Disorgani
zation .42
Family Disorgani
zation .39 Bureaucratization -.32
> $10,000 .35 Migration .29
Migrant .32 Density -.23
Density -.28 < $3,000 -.15
Bureaucratization -.25 < 5 y rs. ed. -.11
Social Mobility -.1 0 Social Mobility -.1 0
Segregation -.0 2 Segregation -.07
113
population, social m obility makes no important independent c o n tri
bution.
Only the emergence of employment of women as an important
fa c to r in determining f e r t i l i t y is. inconsistent with the b iv ariate
analysis of variab les. Its zero-order co rrelatio n with the f e r t i l i t y
ra tio is -.0 9 , suggesting th a t i t is a re la tiv e ly unimportant facto r.
I t is apparent from the large beta c o e ffic ie n t th a t the e f f e c t of
employment of women is being masked in some complex way by other
variables in the urban system.
When segregation is included in the m ultiple regression
equation, the rankings of variables are not importantly a lte re d .
Family disorganization and high education s t i l l tend to be the most
important variables (Table 12). Furthermore, low education and social
m obility are s t i l l re la tiv e ly unimportant. Segregation i t s e l f appears
to make almost no independent contribution to the f e r t i l i t y ra tio .
Apparently, whatever contribution segregation makes is through such
variables as income and education.
Discussion
These data suggest important stru ctu ra l differences between
the to ta l and the nonwhite populations. Although complete in te rp re
ta tio n s of these differences are not ju s ti f ie d , c ertain im plications
can be explored. In the to ta l population, density and migration are
the two most important factors affecting f e r t i l i t y . This implies th a t
cost facto rs and a lte ra tio n s in the age stru ctu re of communities are
of major importance.
114
Density may have re la tiv e ly l i t t l e impact upon nonwhite
f e r t i l i t y because of the concentration of th is population in ghettos.
That i s , th e ir confinement in densely se ttle d central c itie s may
lim it th e ir choice and consequently th e ir d efin itio n of "cost;" when
there are few a ltern ativ es available to a population, the lim itations
imposed by space may not be perceived in the same way as by populations
with extensive choice.
The re la tiv e unimportance of migration may re fle c t the fa c t
th a t nonwhite populations are extremely youthful v is-a-v is to tal
populations. Consequently, the addition of youthful migrants has
l i t t l e impact upon the "familism" of th is community. I t is emphasized
th a t density and migration do a ffe c t nonwhite f e r t i l i t y ; th is
explanation refers to the re la tiv e importance of these variables in
the two populations.
Family disorganization and employment of women emerge as the
most important determinants of nonwhite f e r t i l i t y . Regarding family
disorganization, i t appears th a t the h isto rical emphasis given to th is
variable as a fa c to r in nonwhite f e r t i l i t y is ju s tif ie d . W hen the
family stru ctu re is weakened, the re s u lt is high f e r t i l i t y . In the
to ta l community, th is variable is not only of much less Importance,
i t also has opposite e ffe c ts . This may r e f le c t the re la tiv e ly strong
controls over sexual behavior independent of the family. I t 1s also
possible th a t these controls are strengthened when the family
stru ctu re is weakened.
Also important is the integration of women in the nonwhite
115
population. Although th is is of moderate importance in the to tal
population, i t is a crucial fa c to r in the nonwhite population. The
re la tiv e unimportance of th is facto r in the to ta l community may
indicate th a t the p articip a tio n of. women in jobs outside the home is
not as crucial in th e ir integration into the society as in the
nonwhite population. For nonwhite females such p articip a tio n may
r e f le c t an important break with the family.
Another in te re stin g difference between the two populations is
the e ffe c t of education and income. In the to ta l population, low
income and low education are considerably more important than high
income and high education. In the nonwhite population, exactly the
opposite is tru e. Furthermore, in the nonwhite population, high
income is p o sitiv ely rela ted to f e r t i l i t y . This l a t t e r finding
suggests, as was indicated above, th a t the independent e ffe c t of
increasing income is to ra ise f e r t i l i t y . Increases in the re la tiv e
size of the highly educated group apparently influences f e r t i l i t y
behavior markedly in the nonwhite population. The exact nature of
th is influence cannot be specified with the data in th is study.
CHAPTER VI
S U M M A R Y A N D IMPLICATIONS
Summary
A principal goal of the d isse rta tio n is to explore and analyze
the c h a ra c te ristic s of the c ity which determine f e r t i l i t y . An
e x p lic it assumption is th a t i t is not the c ity per se_which affects
f e r t i l i t y , but dimensions of urban stru ctu re . Data presented in
Chapter IV are consistent with th is assumption. I t does not appear
to be a general concept lik e "urbanism" which determines f e r t i l i t y
levels in c i t i e s , but the various dimensions along which urban
stru ctu res vary.
The supported hypotheses outlined in Chapter III provide
insights into the determinants of urban f e r t i l i t y . Variables included
in these hypotheses were examined in a sp ec ific a lly urban context, and
suggest a s e t of "causes" of urban f e r t i l i t y . I t appears from these
data th a t cost factors and migration are of great importance. Less
important are bureaucratization, integration of women, family d is
organization, and high education. Relatively unimportant are low and
high income, low education, and social m obility.
Hypotheses specific to the nonwhite population of the sample
c itie s were also examined. With a single exception, the e x p lic it
assumption was th a t f e r t i l i t y in the two populations is determined in
116
117
the same way by the same fa c to rs. This assumption was usually
ju s tif ie d . In fa c t, the most general impression about determinants
o f f e r t i l i t y in the two populations is th e ir overall s im ila rity ; in
only one instance does a variable operate in an opposite direction
in the nonwhite as in the to tal population. This exception is family
disorganization, which is positively correlated with f e r t i l i t y in the
nonwhite population and negatively correlated with f e r t i l i t y in the
to tal population.
However, within the context of th is overall sim ila rity there
are differences in emphasis. The most important determinants of non
white f e r t i l i t y , as measured by the magnitudes of correlation and
regression c o e ffic ie n ts, are somewhat d iffe re n t than in the to ta l
population. The most important v aria b les, according to these c r i t e r i a ,
appear to be high education, high income, and family disorganization.
In the m ultiple regression analysis, integration of women also emerges
as one of the most important v ariab les; th is is not the case in the
correlation an aly sis, where i t appears th a t integration of women is
of very l i t t l e importance. Of moderate importance in the nonwhite
population are m igration, density, and bureaucratization.
Implications
Ecological Perspectives
As indicated in Chapter I , the underlying perspective in th is
study is human ecology. F e r tility is analyzed as the dependent
variable in an interdependent system of variables. I t is apparent
from data presented above th a t the ultim ate level of f e r t i l i t y in a
118
given c ity is the outcome of a complex in terp lay of v ariab les. These
variables sometimes work in the same d ire c tio n , and sometimes in
opposite d irec tio n s.
For example, i t was shown th a t migration acts d ire c tly to
increase f e r t i l i t y . However, i t has other in d ire c t e ffe c ts . An
increase in migration also leads to increases in other stru ctu ra l
variables whose e ffe c t is to lower f e r t i l i t y . Migration is po sitiv ely
correlated with education and income, which are negatively correlated
with f e r t i l i t y .
There are numerous other examples of th is idea of a network
of interdependent variab les. I t was shown for the to ta l population
th a t bureaucratization is p art of the path through which education
a ffe c ts f e r t i l i t y , and i t also appears to have in d ire c t effe c ts
through i t s association with income. Bureaucratization also has d ire c t
e ffe c ts upon f e r t i l i t y , effects which are p artly masked by i t s
association with migration.
This approach is consistent with the concept of m ultiple
causation. Although the causal ordering of many of the variables used
in the study are unclear, f e r t i l i t y is c learly a dependent variable.
This variable is caused by a large number of independent v aria b les,
no one of which produces a large amount of change. For example, the
correlation between high education and the f e r t i l i t y ra tio is -.15.
This c o rre la tio n , when interpreted as a slope, indicates th a t a unit
change in high education produces only .15 u n its of change in the
f e r t i l i t y r a tio . If th is co rrelatio n is interpreted in iso la tio n from
other variables in the urban ecosystem, i t might be regarded as
119
unimportant.
However, when a ll variables considered in the study, including
high education, are analyzed together in a m ultiple correlation
equation, the re su lta n t co rrelatio n c o e ffic ie n t is .65. This
indicates th a t a substantial proportion of the variance in the f e r t i l
ity ra tio is being accounted fo r.
Consequently, an important im plication of th is study is th a t
"the" cause of f e r t i l i t y in urban contexts cannot be iso la ted .
Furthermore, there do not appear to be even two or three variables
which act to produce or cause f e r t i l i t y . Rather, the concept of
m ultiple causation seems appropriate.
A fu rth e r im plication of th is ecological perspective concerns
the use of gross categories. When a multidimensional concept lik e
"urbanism" is correlated with f e r t i l i t y , the re s u lt is a strong
re la tio n . However, when the separate dimensions along which urban
stru ctu res may vary are correlated individually with f e r t i l i t y , the
magnitudes of the co rrelatio n s are much sm aller.
This disaggregation has fu rth er im plications for the occasional
reversals in the re la tio n between urbanism and f e r t i l i t y . As was
indicated in Chapter I , urbanism sometimes leads to high f e r t i l i t y . A
potential explanation is th a t certain of the variables considered in
th is study "mix" in such a way as to produce th is anomaly. For
example, i t is possible th a t migration is extensive, while the
integration of women and bureaucratization levels are low. S im ilarly,
i t is probable th a t the per cent of these populations which have less
than fiv e years of education is very high.
120
I t is not argued th a t precisely the same patterns observed
in th is study are operative everywhere. For example, in under
developed countries migration and low education may be po sitiv ely
rath er than inversely rela ted as found in th is study. Sim ilarly,
family disorganization may be positively correlated with f e r t i l i t y in
these environments, fo r the same reasons th a t th is valence was found
in nonwhite populations in th is study. Nonetheless, i t is argued th a t
th is perspective provides a meaningful way of understanding the
occasional reversals in the rela tio n sh ip between f e r t i l i t y and
urbanism. Only future research can untangle the precise nature of
these patterns in p a rtic u la r cu ltu res.
Rural-Urban Differences
Although th is study is confined to urban areas, there are
im plications for rural environments. For example, cost factors are
lik e ly to be important determinants of rural as well as urban
f e r t i l i t y . Thus, as i t becomes increasingly less p ro fitab le to use
the labor of children on modern mechanized farms, th e ir cost increases,
and f e r t i l i t y may decrease. S im ilarly, as the per cent of the rural
population with a college education increases, f e r t i l i t y may decrease.
Both of these predictions are based upon the assumption th a t other
variables do not mask these rela tio n sh ip s.
At the present time, many of the dimensions of urban stru ctu re
considered here are more or le ss confined to urban environments. For
example, i t is unlikely th a t bureaucratic stru ctu res or integration
of women into occupational stru ctu res are important variables in rural
121
areas. However, i t does not then follow th a t such variables are
unimportant in rural environments. Rather, i t is the absence of these
variables which is being responded to . For example, i f women are not
employed outside the home, then th is lack of integration is a pressure
toward higher f e r t i l i t y . Nonetheless, when th is situ a tio n is a
constant one across several rural area s, i t cannot be regarded as a
variable affecting f e r t i l i t y in these areas.
This is a c le a r-c u t issue only when d is tin c tly urban or
d is tin c tly rural areas are considered. Duncan and Reiss (1956) argue
th a t such conditions represent two poles on a continuum. For most
variables they show th a t there is a gradient between these two
extremes. As distance from the center of the c ity increases, propor
tions employed in white c o lla r occupations and income decrease, while
f e r t i l i t y increases. Sim ilar gradients e x ist fo r most other variables
considered in th is study.
This discussion suggests the conclusion th a t there is nothing
unique to the "city" which determines f e r t i l i t y . Rather, certain
classes of variables are important determinants of f e r t i l i t y wherever
they are found. However, the m atter is probably more complex. "Laws"
imply statements of conditions under which a given rela tio n sh ip holds.
The discussion of re la tio n s presented in Chapters II and III
necessarily included only a lim ited statement of such conditions;
these conditions were introduced prim arily as control variab les. For
example, i t was argued th a t bureaucratization is inversely related to
f e r t i l i t y under the condition th a t migration is controlled.
122
There is another p o s s ib ility . The type of environment, rural
versus urban, may in te ra c t to produce d iffe re n t relationships among
variables. For example, rural environments may s e le c t people who are
oriented toward large fam ilies. Such persons, even when employed in
bureaucratic occupations, may s t i l l have many children. In other
words, the re la tio n between bureaucratization may not be the same in
rural as in urban environments.
F e r tility Control
This study also has im plications fo r public policy aimed a t
f e r t i l i t y control. The most general finding is th a t a t the aggregate
level the re la tio n between specific variables and f e r t i l i t y may be
very d iffe re n t than these same relatio n s fo r individuals. For example,
although female education is closely associated with the f e r t i l i t y of
individual women, i t appears to be a re la tiv e ly unimportant facto r a t
the level of c i tie s . This implies th a t a system of f e r t i l i t y control
would not be very effectiv e i f based upon increasing the average level
of female education. Of much greater relevance would be a policy of
increasing the integration of women, perhaps through more effectiv e
laws governing th e ir employment.
S im ilarly, these data indicate the importance of cost facto rs.
Of course, no sensible public policy could be directed a t increasing
the density of a given settlem ent. But "cost" can be manipulated in
other ways. For example, decreases in the exemptions currently given
to heads of households with dependent children, increases in the tax
on items intended for children, e tc.
123
This argument appears "obvious." However, i t is based upon
a clear theoretical ratio n ale and indicates d irection of policy a t
those variables most lik ely to produce an e ffe c t. Furthermore, i t is
homologous to economic p o lic ie s. The underlying assumption in an
increase in in te re s t rates is th a t loans will become s lig h tly more
expensive, resu ltin g in increases in the cost of purchasing such items
as automobiles. Although consumers will continue to buy automobiles,
i t is e x p lic itly assumed th a t on the average consumption of these
items w ill decrease.
Limitations and Suggestions fo r Future Research
There are a number of problems with th is research. For various
reasons certain factors were not considered which could a l t e r the
conclusions reached. I t is possible to consider how these factors
could be accounted fo r in future research. This is done a t the same
time th a t the conceptual framework and methods of analysis are
evaluated.
Causal Order
One of the most frequent problems encountered in the analysis
of hypotheses was determination of the causal ordering of variables.
Sometimes these problems could be p a rtia lly d e a lt with th e o re tic a lly .
For example, i t was argued th a t education precedes bureaucratization,
and therefore th a t the former can legitim ately be considered a cause
of the l a t t e r .
On other occasions i t was not possible to order variables in
124
th is way. Segregation and i t s co rrelates provide a clear example:
does a high degree of segregation cause low levels of income and
education by iso la tin g blacks from the in s titu tio n s and f a c i l i t i e s of
the to ta l population? Do low levels of income and education force
blacks* to acquire housing and jobs in a p a rtic u la r p art of the city?
Another example is migration: Do bureaucratic structures require
constant in te rc ity exchanges of certain types of personnel? Does
migration of these kinds of people lead to higher levels of
bureaucratization?
Such questions are c r i t i c a l . The actual ordering of variables
determines whether or not a given relatio n sh ip with f e r t i l i t y is
"spurious" or is "in terp reted ." For example, i f income is temporally
p rio r to both segregation and f e r t i l i t y , then the association between
the l a t t e r variables is spurious. This would re s u lt in the conclusion
th a t segregation is an unimportant variable in the determination of
f e r t i l i t y . If income intervenes temporally between segregation and
f e r t i l i t y , then an important path through which segregation affects
f e r t i l i t y has been iso la ted .
A related problem is mutual causation. I t is possible th a t
bureaucratization and migration have reciprocal e ffects upon one-
another. As with determination of temporal ordering, a clear
th eo retical ratio n ale is e s s e n tia l. However, certain kinds of data
are also required to evaluate hypotheses of th is kind.
One s ta t is tic a l technique th a t deals with th is problem is the
"lag correlation" ( c f . , Pelz and Andrews, 1964; and Bohrnstedt, 1969).
125
This type of correlation requires data on each of the variables used
in the study fo r a t le a s t one census year preceding 1960. With such
data i t would be possible to co rrelate segregation in 1950 with black
income in 1960, and vice v ersa. I f one correlation is su b stan tially
larg er than the o th er, then in sig h t into the proper causal ordering
is obtained. I f both are of approximately equal magnitude, then
support is given to hypotheses about reciprocal e ffe c ts . I t is
suggested th a t future research incorporate these kinds of data.
Other Variables
This d isse rta tio n is concerned with evaluation, in a s p e c ifi
cally urban context, of hypotheses re la tin g independent variables of
known importance to f e r t i l i t y . The goal is the fu rth er understanding
of the determinants of urban f e r t i l i t y and application of the
ecological perspective in th is context. More ab strac t sociological
variables such "egoism," or "moral in teg ratio n ," were excluded by
th is approach because a c le a r theoretical ratio n ale for th e ir
inclusion was lacking and because data were not available fo r th e ir
measurement. The preliminary analysis performed here should provide
a s u ffic ie n tly firm empirical stru ctu re within which these more
a b strac t concepts can be included once adequate operational d efin itio n s
are developed.
There are also important methodological reasons for including
these variables. An im p lic it assumption underlying the m ultiple
regression analysis and t e s t of b iv ariate hypotheses is th a t all
relevant variables are included. This probably is not the case.
126
Exclusion of relevant variables is a p o te n tia lly serious
problem, because the magnitudes of a given s e t of regression c o e ffic
ien ts depend s p e c ific a lly upon the variables included in the equation
( c f . , Gordon, 1968). I f variables, which are correlated with a given
s e t of independent variables are excluded from a m ultiple regression
equation, spuriously high regression c o efficien ts may re s u lt.
Consequently, i t may be erroneously concluded th a t a p a rtic u la r
variable makes a large independent contribution to f e r t i l i t y .
Analogous dangers e x is t in the in te rp re ta tio n of b iv a riate
hypotheses. I f an unanalyzed variable is correlated with two v a ri
ables included in a hypothesis, the predicted rela tio n sh ip may be
spuriously high or spuriously low.
Of course, these problems are always present. As Blalock
(1961) notes, a t some point the researcher must consider his system
closed. However, i t is suggested th a t the th eo retical and methodo
logical ra tio n a le fo r the analysis of urban f e r t i l i t y can be extended
to include more a b stra c t v ariab les. This requires development of
adequate operational d efin itio n s of these concepts, as well as clear
th eo retical ju s tif ic a tio n fo r th e ir use.
A dditivity Assumption
The perspective developed above s p e c ific a lly requires an
a d d itiv ity assumption. That i s , the assumption th a t u n it increments
in two variables leads to twice the change in a dependent variable as
a u n it change in ju s t one of these v aria b les. However, i t may be th a t
the slopes of the regression lin es d if f e r according to the level of
127
some th ird variable. For example, i t may be th a t income and f e r t i l i t y
are rela ted in d iffe re n t ways under the condition of high education
than they are under the condition of low education.
Such hypotheses require a more sophisticated theory than
proposed here. However, th is and sim ilar studies can be extended to
permit hypotheses about in teractio n e ffe c ts. These studies can
u ti liz e techniques fo r the analysis of in teractio n now being developed
(fo r example, Blalock, 1965).
Lack of Rural Data
As noted above, th is study can only be im p licitly comparative
because data on rural environments are not included. Future studies
ought to include data on such areas to determine whether or not the
facto rs outlined in th is analysis operate in the same way in rural
se ttin g s . I t is possible th a t other variables are of importance, or
th a t they in te ra c t with type of environment so as to produce d iffe re n t
e f f e c ts .
Unit of Study
As indicated in Chapter I I I , central c itie s are used as the
u n it of analysis in th is study. This corresponds to the p o litic a l
boundaries of c i t i e s . Several urban sociologists ( c f ., Davis, 1961)
have argued th a t th is d efin itio n is somewhat a rb itra ry and excludes
the real boundaries of the c ity . I t is possible th a t certain con
clusions would be altered i f urbanized areas, which correspond to the
central c ity and the densely s e ttle d areas contiguous to i t , were
128
used as the u n it of analysis. Future research ought to examine th is
p o ssib ility .
United States Setting
A final problem concerns the focus upon United States c i t i e s .
Such an emphasis greatly lim its the degree to which re su lts can be
generalized. I t may be th a t the re la tio n s found in th is study are
not applicable in other cu ltu res. Consequently, i t is proposed th a t
future research include urban areas in other cu ltu res.
Because of the lim itations summarized in th is sectio n , the
findings of the study are somewhat te n ta tiv e . I t may be th a t more
sophisticated methods of analysis would lead to a lte ra tio n s in the
conclusions reached. Nonetheless, i t is hoped th a t th is research is
a meaningful contribution to human ecology and f e r t i l i t y and th a t
important questions are raised fo r future research.
REFERENCES
Adelman, Irma.
1963. "An Economic Analysis of Population Growth." American
Economic Review 53 (June): 314-339.
Banks, James.
1954 Prosperity and Parenthood. London: Routledge and Kegan Paul.
Becker, Gary.
1960 "An Economic Analysis of F e r ti lit y ." Pp. 209-231 in National
Bureau Committee for Economic Research (ed. ) , Demographic and
Economic Change in Developed Countries. Princeton:
Princeton U niversity Press.
B ell, Wendell.
1958 "Social Choice, Life S tyles, and Suburban Residence." Pp.
225-247 in William Dobriner (e d .), The Suburban Community.
New York: George Putnam.
Beshers, James.
1967 Population Processes in Social Systems. Glencoe: The Free
Press.
Blake, Judith.
1965 "Demographic Science and the Redirection of Population
Policy." Journal of Chronic Diseases 18 (November): 1181-1200.
Blalock, Hubert.
1961 Causal Inferences in Nonexperimental Research. Chapel H ill:
University of North Carolina Press.
1965 "Theory Building and the S ta tis tic a l Concept of Interaction."
American Sociological Review 29 (June): 374-380.
1968 "Theory Building and Causal Inferences." Pp. 155-198 in
Hubert M . Blalock and Ann Blalock (e d s .), Methodology in
Social Research. New York: McGraw-Hill.
Blau, P eter, and Otis D. Duncan.
1964 The American Occupational S tructure. New York: John Wiley
and Sons.
Bogue, Donald.
1969 Principles of Demography. New York: John Wiley and Sons.
129
130
Bohrnstedt, George.
1969 "Observations on the Measurement of Change." Pp. 113-133 in
Edgar F. Borgatta (e d .), Sociological Methodology. San
Francisco: Jossey-Bass.
Bruton, Henry.
1965 Principles of Development Economics. Englewood C liffs:
P rentice-H all.
Collver, Andrew.
1965 Birth Rates in Latin America: New Estimates of H istorical
Trends and Fluctuations. Berkeley: In s titu te of
International Studies.
1968 "Women's Work P articipation and F e r tility in Metropolitan
Areas." Demography 5: 55-60.
Collver, Andrew, and Eleanor Langlois.
1962 "The Female Labor Force in Metropolotan Areas, An Internation
al Comparison." Economic Development and Cultural Change 10
(July): 367-385.
C urtis, Richard, and Elton Jackson.
1962 "Multiple Indicators in Survey Research." American Journal
of Sociology 68 (September): 195-204.
Davis, Kingsley.
1951 The Population of India and Pakistan. Princeton: Princeton
University Press.
1961 "Foreward." Pp. xi-xxii in Jack P. Gibbs (e d .), Urban
Research Methods. New York: D. Van Nostrand.
Driver, Edwin.
1963 D ifferential F e r tility in Central India. Princeton:
Princeton University Press.
Dubois, P hilip.
1957 M ultivariate Correlational Analysis. New York: Harper
Brothers.
Duesenberry, James.
1960 "Comment." Pp. 231-234 in National Bureau of Economic
Research (e d .), Demographic and Economic Change in Developed
Countries. Princeton: Princeton University Press.
Duncan, O tis.
1959 "Human Ecology and Population Studies." Pp. 678-716 in Philip
M. Hauser and Otis D. Duncan (e d s .), The Study of Population.
Chicago: The University of Chicago Press.
1965 "Farm Background and D ifferential F e r tility ." Demography
2: 240-249.
131
Duncan, O tis.
1966 "Path Analysis: Sociological Examples." American Journal of
Sociology 72 (July): 1-16.
Duncan, O tis, and Albert Reiss.
1956 Social C haracteristics of Urban and Rural Communities: 1950.
New York: John Wiley and Sons.
E asterlin , Richard.
1962 The American Baby Boom in H istorical Perspective. New York:
National Bureau of Economic Research.
1965 "Long Swings in United States Demographic and Economic
Growth: Some Findings on the H istorical P attern." Demography
2: 490-507.
1966 "On the Relation of Economic Factors to Recent and Projected
F e r tility Changes." Demography 3: 131-153.
Edin, Karl, and Edward Hutchinson.
1935 Studies of D ifferential F e r tility in Sweden. London:
P. S. King and Son.
Farley, Reynolds.
1965 "The Demographic Rates and Social In s titu tio n s of the
Nineteenth-Century Negro Population: A Stable Analysis."
Demography 2: 386-398.
Eebvay, Maurice.
1959 "Level and Trend of F e r tility by Socio-economic Groups in
France." Population 14 (October-December): 729-739.
Freedman, Ronald.
1961- "The Sociology of Human F e r tility : A Trend Report and
62 Bibliography." Current Sociology 10/11: 35-121.
Freedman, Ronald, Pascal K. Whelpton, and Arthur A. Campbell.
1959 Family Planning, S te r i lit y , and Population Growth. New York:
McGraw-Hi11.
Fried!ander, Stanley, and Morris S ilver.
1967 "A Q uantitative Study of the Determinants of F e r tility
Behavior." Demography 4 (No. 1): 30-70.
G albraith, V irginia, and Dorothy S. Thomas.
1941 "Birth Rates and the Interwar Business Cycles." Journal of
the American S ta tis tic a l Association 36 (December): 465-476.
132
Gibbs, Jack P ., and Walter T. Martin.
1959 "Toward a Theoretical System of Human Ecology." Pacific
Sociological Review 2 (Spring): 29-36.
1962 "Urbanization, Technology, and the Division of Labor:
International P attern s." American Sociological Review 27
(October): 667-678.
Gordon, Robert.
1968 "Issues in M ultiple Regression." American Journal of
Sociology 73 (March): 592-616.
G rab ill, William, Clyde Kiser, and Pascal Whelpton.
1958 The F e r tility of American Women. New York: John Wiley and
Sons.
Hadden, Je ffre y , and Edgar Borgatta.
1965 American C itie s: Their Social C h arac teristics. Chicago:
Rand McNally.
Hamilton, Horace.
1964 "The Negro Leaves the South." Demography 1: 429-443.
Hashmi, Sultan.
1964 "Factors in Urban F e r tility Differences in the United S tates."
Pp. 42-58 in Ernest W . Burgess and Donald J . Bogue (e d s.),
Contributions to Urban Sociology. Chicago: U niversity of
Chicago Press.
Hawley, Amos.
1950 Human Ecology: A Theory of Community Structure. New York:
Ronald.
Heer, David.
1964 " F e r tility Differences Between Indian and Spanish Speaking
Parts of Andean Countries." Population Studies 18 (July):
71-84.
1966 "Economic Development and F e r ti lit y ." Demography 3: 423-444.
Heer, David, and Elsa Turner.
1965 "Areal Differences in Latin American F e r tility ." Population
Studies 19 (March): 279-292.
Hexter, Maurice.
1925 Social Consequences of Business Cycles. New York: Houghton-
M ifflin.
H irschi, T ravis, and Hanan Selvin.
1967 Delinquency Research: An Appraisal of Analytic Methods.
New York: The Free Press.
133
J a ffe , Abram.
1942 "Urbanization and F e r tility ." American Journal of Sociology
48 (July): 48-60.
Keyes, Fenton.
1958 "The Correlation of Social Phenomena with Community Size."
Social Forces 36 (May): 311-315.
Kirk, Dudley.
1960 "The Influence of Business Cycles on Marriage and Birth
Rates." Pp. 241-257 in National Bureau of Economic Research
(e d .), Demographic and Economic Change in Developed Countries.
Princeton: Princeton U niversity Press.
K iser, Clyde.
1942 Group Differences in Urban F e r tility . Baltimore: The
Williams and Wilkins Co.
1958 " F e r tility Trends and D ifferen tials Among Non-whites in the
United S tates." Mil bank Memorial Fund Quarterly 36 (A pril):
149-197.
Labovitz, Sanford.
1963 Technology and the Division of Labor. Unpublished Ph.D.
d is s e rta tio n , the U niversity of Texas a t Austin.
Labovitz, Sanford, and Robert Hagedorn.
1969 Introduction to Social Research. Unpublished manuscript,
Los Angeles.
Lazarsfeld, Paul.
I960 "Problems in Methodology." Pp. 39-78 in Robert K. Merton,
Leonard Broom, and Leonard C o ttre ll, J r . (e d s .), Sociology
Today: Problems and Prospects. New York: Basic Books.
Lorimer, Frank, and Frederick Osborn.
1934 Dynamics of Populations. New York: The Macmillan Co.
M arshall, Harvey.
1968 "Urbanism and the Demographic T ransition." Unpublished
manuscript, Los Angeles.
Miro, Carmen, and Ferdinand Rath.
1965 "Preliminary Findings of Comparative F e r tility Surveys in
Three Latin American C itie s." Mil bank Memorial Fund
Quarterly 43 (October): 36-62.
McNemar, Quinn.
1962 Psychological S ta t is tic s . New York: John Wiley and Sons.
134
Moberg, Sven.
1950 "Marital Status and Family Size Among M atriculated Persons
in Sweden." Population Studies 4 (June): 115-127.
Ogburn, William, and Dorothy Thomas.
1922 "The Influence of the Business Cycle on Certain Social
Conditions." Quarterly Publications of the American
S ta tis tic a l Association 18 (September): 324-340.
Okun, Bernard.
1958 Trends in Birth Rates in the United States Since 1870.
Baltimore: The Johns Hopkins Press.
Pelz, Donald., and Frank Andrews.
1964 "Detecting Causal P rio ritie s in Panel Study Data." American
Sociological Review 29 (December): 836-848.
Petersen, William.
1960 "The Demographic Transition in the Netherlands." American
Sociological Review 25 (June): 334-347.
1961 Population. New York: The Macmillan Co.
Robinson, Warren.
1963 "Urbanization and F e r tility : The Non-Western Experience."
Mil bank Memorial Fund Quarterly 61 (July): 291-308.
Robinson, Warren., and Elizabeth Robinson.
1950 "Rural-Urban F e r tility D ifferen tials in Mexico." American
Sociological Review 25 (February): 77-81.
S irk in , Gerald.
1961 Introduction to Macroeconomic Theory. Homewood, 111.:
Richard D. Irwin.
Spengler, Joseph.
1959 "Economics and Demography." Pp. 791-831 in Philip M . Hauser
and Otis D. Duncan (e d s .), The Study of Population.
Chicago: University of Chicago Press.
Stycos, Joseph.
1968 Human F e r tility in Latin America: Sociological Perspectives.
Ithaca: Cornell University Press.
Stycos, Joseph., and Robert Weller.
1967 "Female Working Roles and F e r ti lit y ." Demography 4: 210-217.
Taueber, Irene.
1958 The Population of Japan. Princeton: Princeton U niversity
Press.
135
Taueber, K arl., and Alma Taueber.
1965 Negroes in C itie s . Chicago: Aldine.
Theodorson, George.
1967 "The Uses of Causation in Sociology." Pp. 131-152 in
Llewellyn Gross (e d .), Sociological Theory. New York:
Harper and Row.
Thomas, Dorothy.
1927 Social Aspects of the Business Cycle. New York: Alfred
Knopf.
1960 "Comment." Pp. 257-260 in National Bureau of Economic
Research (e d .), Demographic and Economic Change in Developed
Countries. Princeton: Princeton U niversity Press.
Thompson, Warren.
1931 Ratios of Children to Women: 1920. Census Monograph 11.
Washington, D.C.: United S tates Government Printing O ffice.
1933 Population Problems. New York: McGraw-Hill.
Tien, Yuen.
1967 "Mobility, Non-Familial A ctivity, and F e r tility ."
Demography 4: 218-227.
United Nations.
1961 "Demographic Aspects of Urbanization in Latin America."
Pp. 91-117 in Philip M . Hauser (e d .), Urbanization in Latin
America. New York: International Documents Service.
United Nations.
1953 The Determinants and Consequences of Population Trends.
United Nations Population Studies. New York: United Nations.
United States Bureau of Census.
1963 United States Census of Population: 1960. Vol. 1,
C h aracteristics of the Population. Washington, D.C.: United
S tates Printing Office.
Weintraub, Robert.
1962 "The Birth Rate and Economic Development; An Empirical
Study." Econometrics 40 (October): 812-817.
Whelpton, P ascal., and Clyde Kiser.
1943 "D ifferential F e r tility Among 41,498 Native White Couples in
Indianapolis." Mil bank Memorial Fund Quarterly 21 (July):
221-280.
1958 "Social and Psychological Factors Affecting F e r tility ."
Milbank Memorial Fund Quarterly 5 (July): 283-329.
136
YelTin.
1955
Z arate,
1967
Seymour.
"Social Mobility and Familism." Unpublished Ph.D. d is s e rta
tio n , Northwestern U niversity.
Alvan.
" F e rtility in Urban Areas of Mexico: Implications for the
Theory of the Demographic T ransition." Demography 4:
363-373.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

S.T.P.: A Simulation Of Treatment Processes

PDF

Referential Dissociation And Response To Stress

PDF

Factors In The Conviction Of Law Violators: The Drinking Driver

PDF

A Sociological Approach To The Etiology Of Female Homosexuality And The Lesbian Social Scene

PDF

A Study Of Factors Related To Police Diversion Of Juveniles: Departmentalpolicy And Structure, Community Attachment, And Professionalization Of Police

PDF

The Effects Of Social Security Programs On Fertility Levels

PDF

Resource allocation among organizations within urban communities

PDF

Social-Psychological Factors In Organizational Attachment

PDF

Status Consistency Among The Clergy

PDF

Economic Differentiation And Social Organization Of Standard Metropolitanareas In The United States: 1950

PDF

Determinants Of Intercounty Migration: California, 1970-1973

PDF

The Effect Of Motor Ability Loss On Cognition And Emotion

PDF

Selected Social Psychological Factors Related To Viewers Of Television Programs

PDF

An Experimental Study Of Factors Influencing The Perception Of Auditory Flutter

PDF

Social Components Of Housing Cost In The Western Metropolis

PDF

Intellect After Lobotomy In Schizophrenia: A Factor-Analytic Study

PDF

The Construction And Empirical Test Of A Theory Based On Selected Variables In Small-Group Interaction

PDF

Some Behavioral Consequences Of Career Success: A Synthesis Of Reward Andbalance Approaches

PDF

Differential Fertility Behavior And Values In Rural And Semi-Urban Costa Rica

PDF

Some Social Factors Affecting The Power Structure And Status Of A Professional Association In Reference To Social Work

Asset Metadata

Creator Marshall, Harvey Huston, Jr. (author)

Core Title Structural Factors Affecting Fertility In Large United States Cities

Degree Doctor of Philosophy

Degree Program Sociology

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag Geography,OAI-PMH Harvest,sociology, demography

Language English

Contributor Digitized by ProQuest (provenance)

Advisor Labovitz, Sanford I. (committee chair), Martin, David W. (committee member), Miller, Jon P. (committee member), Turk, Herman (committee member)

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c18-396026

Unique identifier UC11361116

Identifier 7008533.pdf (filename),usctheses-c18-396026 (legacy record id)

Legacy Identifier 7008533.pdf

Dmrecord 396026

Document Type Dissertation

Rights Marshall, Harvey Huston, Jr.

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the au...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus, Los Angeles, California 90089, USA